mkimia.fst.unair.ac.idmkimia.fst.unair.ac.id/wp-content/uploads/2018/04/polymer-libraries.pdfAdvances...

225Advances in Polymer Science

Editorial Board:A. Abe · A.-C. Albertsson · K. Dušek · W.H. de JeuH.-H. Kausch · S. Kobayashi · K.-S. Lee · L. LeiblerT.E. Long · I. Manners · M. Möller · O. NuykenE.M. Terentjev · M. Vicent · B. VoitG. Wegner · U. Wiesner

Advances in Polymer Science

Recently Published and Forthcoming Volumes

Polymer LibrariesVolume Editors: Meier, M.A.R., Webster, D.C.Vol. 225, 2010

Polymer Membranes/BiomembranesVolume Editors: Meier, W.P., Knoll, W.Vol. 224, 2010

Organic ElectronicsVolume Editors: Meller, G., Grasser, T.Vol. 223, 2010

Inclusion PolymersVolume Editor: Wenz, G.Vol. 222, 2009

Advanced Computer SimulationApproaches for Soft Matter Sciences IIIVolume Editors: Holm, C., Kremer, K.Vol. 221, 2009

Self-Assembled Nanomaterials IINanotubesVolume Editor: Shimizu, T.Vol. 220, 2008

Self-Assembled Nanomaterials INanofibersVolume Editor: Shimizu, T.Vol. 219, 2008

Interfacial Processes and MolecularAggregation of SurfactantsVolume Editor: Narayanan, R.Vol. 218, 2008

New Frontiers in Polymer SynthesisVolume Editor: Kobayashi, S.Vol. 217, 2008

Polymers for Fuel Cells IIVolume Editor: Scherer, G.G.Vol. 216, 2008

Polymers for Fuel Cells IVolume Editor: Scherer, G.G.Vol. 215, 2008

Photoresponsive Polymers IIVolume Editors: Marder, S.R., Lee, K.-S.Vol. 214, 2008

Photoresponsive Polymers IVolume Editors: Marder, S.R., Lee, K.-S.Vol. 213, 2008

PolyfluorenesVolume Editors: Scherf, U., Neher, D.Vol. 212, 2008

Chromatography for Sustainable PolymericMaterialsRenewable, Degradable and RecyclableVolume Editors: Albertsson, A.-C.,Hakkarainen, M.Vol. 211, 2008

Wax Crystal Control · NanocompositesStimuli-Responsive PolymersVol. 210, 2008

Functional Materials and BiomaterialsVol. 209, 2007

Phase-Separated Interpenetrating PolymerNetworksAuthors: Lipatov, Y.S., Alekseeva, T.Vol. 208, 2007

Hydrogen Bonded PolymersVolume Editor: Binder, W.Vol. 207, 2007

Oligomers · Polymer CompositesMolecular ImprintingVol. 206, 2007

Polysaccharides IIVolume Editor: Klemm, D.Vol. 205, 2006

Neodymium Based Ziegler Catalysts –Fundamental ChemistryVolume Editor: Nuyken, O.Vol. 204, 2006

Polymer Libraries

Volume Editors: Michael A.R. MeierDean C. Webster

With contributions by

N. Adams · C.R. Becer · K.L. Beers · M.J. FasolkaM.A.R. Meier · U.S. Schubert · C.M. StaffordD.C. Webster

123

EditorsProf. Dr. Michael A.R. MeierUniversity of PotsdamInstitute of ChemistryLaboratory of Sustainable Organic SynthesisKarl-Liebknecht-Str. 242514476 Golm/[email protected]

Prof. Dean C. WebsterDepartment of Coatings

and Polymeric MaterialsNorth Dakota State UniversityPO Box 6050, Dept 2760Fargo, ND 58108, [email protected]

ISSN 0065-3195 e-ISSN 1436-5030ISBN 978-3-642-00169-7 e-ISBN 978-3-642-00170-3DOI 10.1007/978-3-642-00170-3Springer Heidelberg Dordrecht London New York

Library of Congress Control Number: 2010926038

c© Springer-Verlag Berlin Heidelberg 2010This work is subject to copyright. All rights are reserved, whether the whole or part of the material isconcerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publicationor parts thereof is permitted only under the provisions of the German Copyright Law of September 9,1965, in its current version, and permission for use must always be obtained from Springer. Violationsare liable to prosecution under the German Copyright Law.The use of general descriptive names, registered names, trademarks, etc. in this publication does notimply, even in the absence of a specific statement, that such names are exempt from the relevant protectivelaws and regulations and therefore free for general use.

Cover design: WMXDesign GmbH, Heidelberg

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Volume Editors

Prof. Dr. Michael A.R. Meier

University of PotsdamInstitute of ChemistryLaboratory of Sustainable Organic SynthesisKarl-Liebknecht-Str. 242514476 Golm/[email protected]

Prof. Dean C. Webster

Department of Coatingsand Polymeric Materials

North Dakota State UniversityPO Box 6050, Dept 2760Fargo, ND 58108, [email protected]

Editorial BoardProf. Akihiro Abe

Department of Industrial ChemistryTokyo Institute of Polytechnics1583 Iiyama, Atsugi-shi 243-02, [email protected]

Prof. A.-C. Albertsson

Department of Polymer TechnologyThe Royal Institute of Technology10044 Stockholm, [email protected]

Prof. Karel Dušek

Institute of Macromolecular Chemistry,CzechAcademy of Sciences of the Czech RepublicHeyrovský Sq. 216206 Prague 6, Czech [email protected]

Prof. Dr. Wim H. de Jeu

Polymer Science and EngineeringUniversity of Massachusetts120 Governors DriveAmherst MA 01003, [email protected]

Prof. Hans-Henning Kausch

Ecole Polytechnique Fédérale de LausanneScience de BaseStation 61015 Lausanne, [email protected]

Prof. Shiro Kobayashi

R & D Center for Bio-based MaterialsKyoto Institute of TechnologyMatsugasaki, Sakyo-kuKyoto 606-8585, [email protected]

Prof. Kwang-Sup Lee

Department of Advanced MaterialsHannam University561-6 Jeonmin-DongYuseong-Gu 305-811Daejeon, South [email protected]

Prof. L. Leibler

Matière Molle et ChimieEcole Supérieure de Physiqueet Chimie Industrielles (ESPCI)10 rue Vauquelin75231 Paris Cedex 05, [email protected]

vi Editorial Board

Prof. Timothy E. Long

Department of Chemistryand Research InstituteVirginia Tech2110 Hahn Hall (0344)Blacksburg, VA 24061, [email protected]

Prof. Ian Manners

School of ChemistryUniversity of BristolCantock’s CloseBS8 1TS Bristol, [email protected]

Prof. Martin Möller

Deutsches Wollforschungsinstitutan der RWTH Aachen e.V.Pauwelsstraße 852056 Aachen, [email protected]

Prof. Oskar Nuyken

Lehrstuhl für Makromolekulare StoffeTU MünchenLichtenbergstr. 485747 Garching, [email protected]

Prof. E. M. Terentjev

Cavendish LaboratoryMadingley RoadCambridge CB 3 OHE, [email protected]

Maria Jesus Vicent, PhDCentro de Investigacion Principe FelipeMedicinal Chemistry UnitPolymer Therapeutics LaboratoryAv. Autopista del Saler, 1646012 Valencia, [email protected]

Prof. Brigitte Voit

Institut für Polymerforschung DresdenHohe Straße 601069 Dresden, [email protected]

Prof. Gerhard Wegner

Max-Planck-Institutfür PolymerforschungAckermannweg 1055128 Mainz, [email protected]

Prof. Ulrich Wiesner

Materials Science & EngineeringCornell University329 Bard HallIthaca, NY 14853, [email protected]

Advances in Polymer SciencesAlso Available Electronically

Advances in Polymer Sciences is included in Springer’s eBook package Chemistryand Materials Science. If a library does not opt for the whole package the bookseries may be bought on a subscription basis. Also, all back volumes are availableelectronically.

For all customers who have a standing order to the print version of Advances inPolymer Sciences, we offer the electronic version via SpringerLink free of charge.

If you do not have access, you can still view the table of contents of each volumeand the abstract of each article by going to the SpringerLink homepage, clickingon “Browse by Online Libraries”, then “Chemical Sciences”, and finally chooseAdvances in Polymer Science.

You will find information about the

– Editorial Board– Aims and Scope– Instructions for Authors– Sample Contribution

at springer.com using the search function by typing in Advances in PolymerSciences.

Color figures are published in full color in the electronic version on SpringerLink.

viii Advances in Polymer Sciences Also Available Electronically

Aims and Scope

The series Advances in Polymer Science presents critical reviews of the presentand future trends in polymer and biopolymer science including chemistry, physicalchemistry, physics and material science. It is addressed to all scientists at universi-ties and in industry who wish to keep abreast of advances in the topics covered.

Review articles for the topical volumes are invited by the volume editors. As arule, single contributions are also specially commissioned. The editors and pub-lishers will, however, always be pleased to receive suggestions and supplementaryinformation. Papers are accepted for Advances in Polymer Science in English.

In references Advances in Polymer Sciences is abbreviated as Adv. Polym. Sci.and is cited as a journal.

Special volumes are edited by well known guest editors who invite reputed authorsfor the review articles in their volumes.

Impact Factor in 2008: 6.802; Section “Polymer Science”: Rank 2 of 73

Preface

This is truly an exciting time to be in the field of polymer science. Advances inpolymerization methods are providing polymer scientists with the ability to specifyand control polymer composition, structure, architecture, and molecular weight toa degree that was not possible just a decade ago. This, in turn, is resulting in manynovel application possibilities of polymers ranging from drug delivery systems andnanolithography to stimuli-responsive materials and many others. In addition, manyof the application areas of polymers – such as coatings, adhesives, thermoplastics,composites, and personal care – are also taking advantage of the ability to designpolymers during their development efforts. Not to forget, many of these applicationsof polymers involve mixing polymers with solvents, catalysts, colorants, and manyother ingredients to prepare a formulated product.

However, the tuning of polymer composition and structure as well as polymerformulations to optimize the final performance properties can be challenging, es-pecially since in many cases several interacting variables need to be optimizedsimultaneously. This is where the methodologies and techniques of combinatorialand high-throughput experimentation to synthesize and characterize polymer li-braries can be an invaluable approach. Simply put, a polymer library is a collectionof multiple polymer samples having a systematic variation in one or more variablesrelated to composition, structure, or process. Various methods and strategies havebeen explored to efficiently prepare a large number of polymer samples and alsoto screen these samples for key properties of interest. In this way, a broad range ofcompositions can be prepared and evaluated in a similar time frame required to pre-pare one or two samples, significantly increasing the efficiency of the experimentalprocess. In addition, because the variable space is explored more thoroughly andin more detail than when using conventional laboratory methods, often materialshaving a unique combination of properties are identified.

While the use of these methods can be shown to be of benefit to a large number ofpolymer research programs, the widespread implementation of these concepts hasnot been realized. Thus, we would encourage those working in complex polymersystems to carefully consider the examples provided in this volume and identifyhow these could be implemented in their research work.

In Chap. 1, we provide an introduction to the strategies that have been reportedfor the preparation and characterization of polymer libraries and then highlight

ix

x Preface

a few selected examples where polymer libraries have been effectively used toidentify novel materials. In Chap. 2, Becer and Schubert describe the preparationof polymers using controlled/living polymerization methods. Automated reactorshave been used both to optimize the synthetic conditions and for preparing librariesof novel block copolymers. Next, Fasolka, Stafford, and Beers describe strategiesused to study the interfaces of polymer systems using a gradient combinatorial ap-proach. In the gradient approach, a single physical sample is prepared that has asystematic change in properties such as composition, thickness, surface energy, etc.A number of truly unique and creative methods have been developed to preparethe samples and characterize the gradient libraries for properties such as adhesion,surface energy, modulus, and so on. Finally, one of the challenges in the use of com-binatorial and high-throughput methods is in the analysis and modeling of the dataobtained. In Chap. 4, Adams discusses various approaches and especially the chal-lenges involved in the modeling of the polymer data which may be generated usingcombinatorial and high-throughput experiments.

While providing a compendium of work done in the past, our primary aim isthat this volume will provide inspiration and motivation for polymer scientists toemploy combinatorial and high-throughput methods in their research efforts andgenerate even greater and novel discoveries from their research work.

In addition, we would like to thank all those who have contributed to this vol-ume to make it a success: C. Remzi Becer, Ulrich S. Schubert, Michael J. Fasolka,Christopher M. Stafford, Kathryn L. Beers, and Nico Adams. Without your excel-lent contributions, this volume would not have been a reality.

Potsdam, Germany Michael A.R. MeierFargo, ND, USA Dean C. WebsterSpring 2010

Contents

Polymer Libraries: Preparation and Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Dean C. Webster and Michael A.R. Meier

Parallel Optimization and High-Throughput Preparationof Well-Defined Copolymer Libraries UsingControlled/“Living” Polymerization Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17C. Remzi Becer and Ulrich S. Schubert

Gradient and Microfluidic Library Approaches to PolymerInterfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63Michael J. Fasolka, Christopher M. Stafford, and Kathryn L. Beers

Polymer Informatics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .107Nico Adams

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .151

xi

Adv Polym Sci (2010) 225: 1–15DOI:10.1007/12_2009_15c© Springer-Verlag Berlin Heidelberg 2009

Published online: 22 October 2009

Polymer Libraries: Preparationand Applications

Dean C. Webster and Michael A.R. Meier

Abstract Polymer libraries offer straightforward opportunities for the investigationof structure–property relationships and for a more thorough understanding of cer-tain research problems. Furthermore, if combined with high-throughput methodsfor their preparation as well as screening, they offer the additional advantage oftime savings and/or the reduction of experimental efforts. Thus, the herein discussedmethods of polymer library preparation and selected literature examples of polymerlibraries describe efficient and state-of-the-art methods to tackle difficult researchchallenges in polymer and materials science.

Keywords Combinatorial materials research · High-throughput screening · Librarypreparation · Polymer library · Property screening

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Polymer Library Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Selected Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

D.C. Webster (�)Coatings and Polymeric Materials, North Dakota State University, 1735 NDSU Research ParkDrive, Fargo, ND 58102, USAemail:[email protected]

M.A.R. MeierUniversity of Potsdam, Institute of Chemistry, Karl-Liebknecht-Str. 24-25, 14476 Golm, Germanyemail:[email protected]

email: [email protected]

email: [email protected]

2 D.C. Webster and M.A.R. Meier

1 Introduction

Polymers are highly tailorable materials and polymers having unique combinationsof properties or can perform a specific function (e.g. drug delivery) are desirable.In addition to variations in composition and molecular weight, different polymerarchitectures, such as block, graft, star, dendrimer, etc., are also possible. Identifyinga specific polymer that has the desired properties can be a challenging task due tothe large number of variations possible.

Therefore, polymer libraries, in combination with high-throughput screeningtechniques, are highly useful tools for the evaluation of (quantitative) structure–property relationships and/or the identification of “hits” of certain desired propertiesof the evaluated materials. These tools help researchers to understand their researchproblems more thoroughly by, e.g., finding optimal process conditions or productperformance within a reduced amount of time and/or experimental effort.

After the introduction of combinatorial and high-throughput approaches inpharmaceutical and catalysis research programs, these methods also became avail-able to the polymer/materials scientist at the beginning of this new century [1–3].Therefore, new and specially adopted preparation and high-throughput screeningtechniques had to be developed, taking the requirements of the fields into account[4]. Examples include parallel synthetic equipment that can handle highly viscouspolymer melts and solutions as well as screening techniques for polymer molecularweights and molecular weight distributions. Only this development made it possibleto prepare and screen polymer libraries within a reasonable amount of time, openingthe possibility to address scientific questions that would otherwise be difficult totackle.

In order to find the desired hits and/or structure–property relationships, the de-sign of a polymer library as well as the availability of suitable screening methodsare crucial. The researchers have to ask themselves which experimental factors willhave an influence and what are reasonable ranges for these factors to be tested. Af-ter this screening process, a further optimization of the screening outcome mightbe necessary and, finally, a model might be developed and tested for its robustness.Traditionally, such optimizations are performed stepwise, one parameter at a time.Unfortunately, this approach can lead to results that are far from the optimum (com-pare Fig. 1, left), since interaction between the investigated factors are most likelynot identified. The advantage of high-throughput approaches on the other hand is thepossibility to screen the complete parameter set, making the identification of hits aswell as optimal process parameters easier (Fig. 1, right).

Before setting up the experiments, or preparing a library, a suitable experimentaldesign has to be chosen [5, 6]. One of the most commonly used designs for polymerlibraries are still the statistical full factorial designs and fractional factorial designs.Both have in common the systematic variation of experimental factors in a set ofdiscrete levels, whereby the latter consist of a carefully chosen subset (fraction) ofthe full factorial design. Increasing the amount of simultaneously investigated pa-rameters exponentially increases the experimental as well as data handling efforts

Polymer Libraries: Preparation and Applications 3

Fig. 1 Classical step-by-step optimization (left) compared to simultaneous screening of all param-eters (right)

and can, at a certain point, not even be tackled with high-throughput approaches. Forthese problems, design-of-experiments approaches that utilize statistical experimen-tal designs and allow for the reduction of the number of experiments to be performedwithout compromising the information content of the generated data have to be ap-plied [5, 6].

Within the following sections we will give a brief overview of the available high-throughput methods for the preparation and screening of polymer libraries and thenfocus our discussion on well recognized literature examples of polymer libraries.

2 Polymer Library Preparation

Polymer libraries are generally not prepared for their own sake, but rather in orderto explore some key property of the investigated materials. The technique used toprepare a given polymer library is often dictated by the method(s) to be used toscreen or characterize the library compositions for the key properties of interest.

Concepts for the preparation of polymer libraries have followed two generalpathways. One method involves the preparation of a single specimen wherein inde-pendent variables are varied spatially across the sample dimensions. The use of thistechnique for polymer systems has been led by researchers at the National Instituteof Standards and Technology (NIST) in the USA. For example, the phase behaviorof binary polymer blends has been studied as a function of composition and temper-ature by preparing a single sample having varying polymer blend composition alongone spatial dimension and placing the sample on a temperature stage having a gra-dient in the orthogonal direction [7]. Illustrated in Fig. 2, the composition gradientwas prepared by slowly filling a syringe from a stirred reservoir of the first polymerwhile a solution of the second polymer was added, creating a change in compositionover time.


Fig. 2 Formation of polymer blend gradient film and phase behavior. Reprinted with permissionfrom [7]. (Copyright 2000 American Chemical Society)

Then the content of the syringe was dispensed in a strip on a glass slide and adoctor blade was used to spread the solution in the orthogonal direction. The phasebehavior of the blend can be determined directly by visual inspection of the sample(Fig. 2, right). Similarly, it was also possible to create gradients in both polymerblend composition and film thickness by accelerating the movement of the doctorblade when spreading the blend solution [8].

The group of Genzer et al. have used surface-initiated polymerization to createsamples having gradients in surface grafting density [9–11] and molecular weight[12]. Surfaces of grafted block copolymers having orthogonal variation in the indi-vidual block lengths have also been prepared [13, 14].

The Beers group at NIST have demonstrated the preparation of surface-graftedcopolymers having a compositional gradient by filling a narrow channel with a gra-dient in monomer composition followed by atom-transfer radical polymerization(ATRP) using a surface-grafted initiator (Fig. 3) [15]. This results in a gradient ofstatistical copolymers having a systematic change in composition from one end ofthe sample to the other. Block copolymer brush gradients could be synthesized usinga two-step technique [16]. The first block was polymerized using surface-initiatedATRP. Then using the first block as macroinitiator, the second block was polymer-ized to have a gradient in block length.

To study the effects of composition on the photopolymerization behavior ofacrylates, the group of Bowman et al. prepared gradient libraries where acrylatecomposition was varied in one dimension and light exposure was varied using amoveable shutter in the orthogonal direction [17–21]. An FTIR microscope wasused to characterize the conversion across the samples and the data from multiplelibraries was then used to derive kinetic models for the photopolymerization.

While the gradient approach is attractive for studying many phenomena, formany studies it is desired to prepare a library of discrete (individual) polymersamples having a systematic variation in composition, molecular weight, crosslinkdensity, or architecture. Using discrete samples frees one from the two (or possiblythree) spatial dimensions of the gradient library, allowing for experimental designshaving three, four, or more variables.


Fig. 3 Preparation of surface-grafted polymer brush composition gradient. Reprinted with per-mission from [15]. (Copyright 2006 Wiley-VCH Verlag GmbH & Co)

One approach to preparing polymer samples for a combinatorial study couldinvolve using conventional laboratory synthesis methods to prepare the desired poly-mers one at a time. There are many studies reported in the literature where a seriesof polymers having some systematic variation in composition or other propertywere prepared and characterized. However, this approach has limitations in timeand resources and becomes unattractive when the synthesis of large numbers ofpolymers is required.

If the chemistry is amenable, it is possible to synthesize a large number of smallsamples of polymers by simply mixing the ingredients in either small vials or mul-tiwell plates. For example, Brocchini et al. prepared a library of 112 polymers bymixing the monomers in individual vials which were placed in a water bath [22].Akinc et al. synthesized a library of 24 unique poly(β -amino esters) via the conju-gate addition of acrylates and amines by mixing the monomers in sample vials fittedwith stir bars [23]. To speed up this process, a liquid handling robot can be used todispense the raw materials into an array of vials [24].

Another approach to the preparation of polymer libraries is to conduct individ-ual polymer synthesis reactions in parallel in small individual reactors. To speedup the process, automated parallel synthesizers designed for use in combinatorialchemistry have been adapted and reactor systems specifically designed for polymersynthesis have also been commercialized (Fig. 4). These reactor systems can auto-mate many of the steps needed to prepare a library of polymers, including dispensingof monomers and other reagents according to the desired recipe, controlling theheating and cooling steps and performing the required purification steps.


Fig. 4 Illustration of (a) Chemspeed automated synthesizer and (b) Symyx batch polymerizationsystem for synthesis of discrete polymer libraries. Reprinted with permission from [90]. (Copyright2007 Taylor & Francis Group, http://www.informaworld.com)

The group of Schubert et al. has demonstrated the utility of using automatedchemical synthesizers for carrying out many different types of polymerizations in-cluding controlled radical polymerizations [25–33], cationic ring-opening polymer-ization [34–38], and anionic polymerization [39, 40]. In addition, block copolymers[41–46] and supramolecular polymers [47–49] have also been synthesized. TheWebster group has used a simple batch polymerization system to synthesize func-tional siloxane and siloxane-polycaprolactone block copolymers via ring-openingpolymerization [50, 51], and also to carry out conventional and controlled freeradical polymerization [52–55]. Rojas et al. have reported that it is challenging toconduct step-growth polymerization to high and reproducible molecular weight us-ing an automated reactor system since, in order to obtain high molecular weightpolymers when two or more monomers are used, equivalent stoichiometry betweenthe monomers, and therefore precise dispensing of the monomers into the individualreactors, is required [56]. Liquid dispensing of the monomers in solution was foundto be more precise than dispensing the solid powders. An alternative approach is toconvert the polymerization from a step-growth (condensation) reaction of monomersto an entropically-driven ring-opening polymerization [57].

A limitation of these simple reactor systems is that agitation is typically by vor-texing or magnetic stirring, thus polymerizations have to be carried out in solutionat relatively low viscosity. Newer reactor systems have been designed that employmechanical mixing, allowing for conducting polymerizations at higher viscosity in-cluding multi-step processes [58]. Some of these automated synthesizers can beprogrammed to withdraw samples periodically during the polymerization reactionfor further analysis, so the course of the reaction can be followed and the reactionkinetics evaluated [59].


Following polymer synthesis, in many cases it is necessary to convert the polymerinto a solid film for property screening. Solutions of thermoplastic polymers can bedeposited directly onto the required substrate and the solvent evaporated to leave thepolymer film. Libraries of thermoset polymers are prepared by dispensing and mix-ing the required components, which can include polymers, crosslinkers, solvents,catalysts, etc., using an automated dispensing and mixing system. The mixtures arethen deposited on an appropriate substrate, usually in an array format, followed bythe curing of the library of samples. A number of methods can be used to deposit theformulation libraries including using a liquid handling pipette to deposit the sam-ples into the wells of a microtiter plate or other multiwell substrate [60, 61], ink-jetprinting [62–65], microcontact printing [66–68], or using an automated device fordepositing and spreading the materials on a substrate [69–71].

3 Selected Examples

In recent years, many examples of polymer libraries applying high-throughputexperimentation concepts for the fast and reliable determination of structure–property relationships were reported in the literature.

Kohn et al. were probably the first to make use excessively of these concepts andreported in 1997 on a “combinatorial approach for polymer design” [22]. Therefore,14 tyrosine-derived diphenols and eight diacids were reacted with each other in upto 32 parallel polymerization reactions to obtain a 112-membered library of strictlyalternating A–B type copolymers with predictable and systematic material propertyvariations. These polymerizations were conducted in separate reactions vessels ina water bath in a 0.2 g scale yielding enough material after purification for the es-tablishment basic material properties as well as certain biological properties. Forinstance, it was shown that the glass transition temperature (Tg) as well as the air–water contact angle of the polymers increased as the number of carbon or oxygenatoms in the polymer backbone and pendent chain decreased in a defined fashion tomention only a few of the found structure–property correlations (Fig. 5).

Moreover, a linear correlation was obtained between cell proliferation and air–water contact angles when polymers having an identical backbone structure butdifferent pendent chains were grouped together. In general, cell proliferation sig-nificantly decreased as the polymer surface became more hydrophobic. In contrast,for those polymers having oxygen-containing diacids in the backbone, cell prolifer-ation was far less sensitive to surface hydrophobicity. In fact, all polymers havingoxygen-containing diacids in their backbone were uniformly good fibroblast growthsubstrates irrespective of their air–water contact angle. In subsequent investiga-tions, models for both protein adsorption onto and cellular response to polymericsurfaces were derived from the discussed 112-membered library using computeddescriptors that are only based on the polymer structures and their glass transitiontemperatures (Tg) [72]. Finally, a variety of other biologically important parame-ters, such as gene expression levels or protein adsorption, were evaluated for this


Fig. 5 Glass transition temperature variation within a library of 112 copolymers. Reprinted withpermission from [22]. (Copyright 1997 American Chemical Society)

polymer library and it was attempted to correlate the outcome of these tests to thechemical structure of the investigated polymers [73–75]. These examples clearlydemonstrate that a designed (targeted) library of polymeric materials is a very use-ful tool to evaluate structure–property relationships and to develop computationalmodels. The knowledge obtained can subsequently be applied for the preparation ofmaterials with certain designed properties with a reduced amount of effort and time[76].

Later, several authors adopted this general library preparation technique to theirspecific needs and reported on the synthesis of different polymer libraries preparedvia step-growth polymerization techniques [77–79]. For instance, Candida antarc-tica lipase in acetonitrile was shown to catalyze efficiently the polycondensation ofa variety of diol monomers, including aliphatic and aromatic diols, as well as car-bohydrates, nucleic acids, and a natural steroid diol, with straight chain diesters toform a library of high molecular weight polymers [79]. Moreover, a linearly varyingcompositional library of 100 different biodegradable polyanhydride random copoly-mers was prepared via polycondensation [77]. It was argued that these materials arepromising carriers for controlled drug delivery and indeed the authors could showthat the rate of release of a model dye could be correlated to the copolymer compo-sition.

In an equally distinguished example, Langer et al. demonstrated the synthe-sis of a 140-membered library of degradable polymers from diacrylate and aminemonomers (compare Fig. 6) that were polymerized via aza-Michael addition chem-istry [80].

The library was screened for DNA-complexing materials as well as gene de-livery vectors, revealing several new materials that were able to condense DNA


Fig. 6 Monomers used for the construction of a 140-membered library of degradable polymers.Structures redrawn from [80]

into small enough structures to be internalized by cells and releasing the DNAin a transcriptionally active form. Further investigation of this library by in vitrotransfections screening methods identified two candidates with very high transfec-tion levels and revealed, due to this highly systematic approach, first correlationsbetween the chemical structure of the polymers and their performance [81]. Subse-quently, the synthesis and screening of a library of 2,350 structurally unique, degrad-able, cationic polymers from a larger subset of similar monomers (compare Fig. 6)became feasible by using automated fluid-handling systems [24]. In particular, ahigh-throughput, cell-based screening method could thus identify 46 new poly-mers that transfect with a higher efficiency than conventional nonviral deliverysystems such as poly(ethyleneimine). Last, but not least, Langer et al. reported thein vitro screening of a 500-membered poly(β -amino esters) library for transfec-tion efficiency and cytotoxicity [82]. Some vectors surpassed the best commerciallyavailable nonviral vectors for in vitro and in vivo gene transfer, and it was observedthat the direct administration of one of these poly(β -amino ester)s complexed toDNA encoding the toxin DT-A could effectively inhibit tumor growth in mice.

In the area of conjugated polymers, Lavastre et al. reported a high-throughput ap-proach for the preparation and screening of poly(arylene ethynylene)s [83]. There-fore, Pd-catalyzed carbon–carbon coupling reactions between 12 dihalogenatedand 8 diethynyl monomers were performed in parallel, yielding a 96-memberedpolymer library that was screened for its fluorescent properties in solution as wellas in thin films utilizing plate reader technology. The authors conclude that theirHTE approach was successfully applied for the fast discovery of potential new can-didates for OLEDs and led to the detection of polymers showing a red, green, orblue solid-sate fluorescence [83].


Moreover, libraries of dendrimers and other branched polymer architectureswere successfully prepared and investigated. In 2001 Hawker et al. presenteda strategy to prepare multiarm star polymers using nitroxide-mediated “living”radical polymerization [84]. Therefore, a macroinitiator together with a mono-functional and a bifunctional monomer were reacted resulting in the formation ofcross-linked moieties with random spacers of the monofunctional monomer, whicheffectively knit together the polymeric arms of the macroinitiator leading to for-mation of soluble star polymers. Since too many parameters had to be investigatedfor this polymerization, only a high-throughput approach enabled the researchers toachieve their goal [84]. Later, the same group reported a library of highly branched,3-dimensional, dendron functional core cross-linked star polymers via a similar ap-proach [85]. Moreover, dendrimer libraries were prepared via click-chemistry [86]as well as via classical approaches [87]. Especially the click-chemistry approachled to a library of functionalized dendritic macromolecules in extremely high yieldsusing no protecting group strategies and only minimal purification steps [86]. There-fore, this strategy presents a significant advance compared to traditional approachesnot only for the synthesis of dendrimer libraries but polymer libraries in general.Last, but not least, a small library of star-shaped block copolymers was shown tobehave as unimolecular micelles and to transfer guest molecules from a water to achloroform phase [45]. This behavior was screened using plate reader technologyand the encapsulation behavior could be correlated to the polymer architecture. Ata later stage, it could be shown that these polymers encapsulated a large variety ofdifferent guest molecules [46], they were able to stabilize metal nano-particles thatwere potent catalysts for C–C coupling reactions [88], and the encapsulated guestmolecules could also be transported within these nano-carriers [89].

High-throughput approaches are frequently applied to the synthesis and evalu-ation of coating libraries [90]. Representative examples include, for instance, thepreparation of acrylate based coatings and their subsequent evaluation by high-throughput screening techniques [91]. Thus, 48-element coatings libraries wereprepared as 8× 6 arrays and evaluated for their abrasion resistance applying self-developed test methods leading to an important productivity improvement of at least10 times over a conventional coating development process [91]. Moreover, Web-ster et al. have worked on coating libraries with reduced adhesion for applicationsas anti-fouling coatings [52, 92, 93]. For instance, an acrylic polyol library wassynthesized using batch solution polymerization of the monomers using the threemonomers butyl methacrylate, n-butyl acrylate, and 2-hydroxyethyl acrylate to ob-tain polyols of varying compositions [52]. The resulting 24-membered polyol librarywas characterized using high-throughputgel permeation chromatography and differ-ential scanning calorimetry (DSC). Subsequently, this library was formulated intosiloxane-polyurethane coatings and the resulting coatings were tested for pseudo-barnacle adhesion revealing that most of the investigated materials showed very lowadhesion [52]. Later, the same authors reported on the development of an automatedimaging software tool that quantifies bacterial and algal percent coverage on coatingarrays and argued that surface coverage is a highly relevant parameter when down-selecting coatings on the basis of their fouling-release potential [92]. The presented


screening method, in combination with other high-throughput screening techniques,would ultimately allow a reduction of 200–300 coating formulations to only approx-imately 10 formulations that need to be subjected to full ocean testing [92]. Thus,these high-throughput coating testing methods not only save time, but also signifi-cantly reduce the time required and the expenses necessary for full evaluation of acertain class of materials.

More specific libraries, such as polymer thin film (gradient) libraries or librariesprepared via controlled polymerization techniques, are not part of this overview, butwill be discussed in detail in other chapters of this book [94, 95].

References

1. Webster DC (2008) Combinatorial and high-throughput methods in macromolecular materialsresearch and development. Macromol Chem Phys 209:237–246

2. Hoogenboom R, Meier MAR, Schubert US (2003) Combinatorial methods, automated syn-thesis and high-throughput screening in polymer research: past and present. Macromol RapidCommun 24:15–32

3. Meier MAR, Hoogenboom R, Schubert US (2004) Combinatorial methods, automated syn-thesis and high-throughput screening in polymer research: the evolution continues. MacromolRapid Commun 25:21–33

4. Schmatloch S, Meier MAR, Schubert US (2003) Instrumentation for combinatorial and high-throughput polymer research: a short overview. Macromol Rapid Commun 24:33–46

5. Cawse JN (2001) Experimental strategies for combinatorial and high-throughput materials de-velopment. Acc Chem Res 34:213–221

6. Harmon L (2003) Experiment planning for combinatorial materials discovery. J Mater Sci38:4479–4485

7. Meredith JC, Karim A, Amis EJ (2000) High-throughput measurement of polymer blend phasebehavior. Macromolecules 33:5760–5762

8. Meredith JC, Smith AP, Karim A, Amis EJ (2000) Combinatorial materials science for polymerthin-film dewetting. Macromolecules 33:9747–9756

9. Wu T, Efimenko K, Vlcek P, Subr V, Genzer J (2003) Formation and properties of anchoredpolymers with a gradual variation of grafting densities on flat substrates. Macromolecules36:2448–2453

10. Wu T, Tomlinson M, Efimenko K, Genzer J (2003) A combinatorial approach to surface an-chored polymers. J Mater Sci 38:4471–4477

11. Wu T, Efimenko K, Genzer J (2002) Combinatorial study of the mushroom-to-brush crossoverin surface anchored polyacrylamide. J Am Chem Soc 124:9394–9395

12. Tomlinson MR, Genzer J (2003) Formation of grafted macromolecular assemblies with a grad-ual variation of molecular weight on solid substrates. Macromolecules 36:3449–3451

13. Bhat RR, Tomlinson MR, Genzer J (2005) Orthogonal surface-grafted polymer gradients: aversatile combinatorial platform. J Polym Sci Part B Polym Phys 43:3384–3394

14. Tomlinson MR, Genzer J (2005) Evolution of surface morphologies in multivariant assemb-lies of surface-tethered diblock copolymers after selective solvent treatment. Langmuir21:11552–11555

15. Xu C, Barnes SE, Wu T, Fischer DA, DeLongchamp DM, Batteas JD, Beers KL (2006)Solution and surface composition gradients via microfluidic confinement: fabrication of astatistical-copolymer-brush composition gradient. Adv Mater 18:1427–1430

16. Xu C, Wu T, Batteas JD, Drain CM, Beers KL, Fasolka MJ (2006) Surface-grafted blockcopolymer gradients: effect of block length on solvent response. Appl Surf Sci 252:2529–2534


17. Johnson PM, Reynolds TB, Stansbury JW, Bowman CN (2005) High throughput kineticanalysis of photopolymer conversion using composition and exposure time gradients. Poly-mer 46:3300–3306

18. Johnson PM, Stansbury JW, Bowman CN (2008) High-throughput kinetic analysis of acrylateand thiol-ene photopolymerization using temperature and exposure time gradients. J PolymSci Part A Polym Chem 46:1502–1509

19. Johnson PM, Stansbury JW, Bowman CN (2007) Photopolymer kinetics using light intensitygradients in high-throughput conversion analysis. Polymer 48:6319–6324

20. Johnson PM, Stansbury JW, Bowman CN (2007) Alkyl chain length effects on copolymeriza-tion kinetics of a monoacrylate with hexanediol diacrylate. J Comb Chem 9:1149–1156

21. Johnson PM, Stansbury JW, Bowman CN (2008) Kinetic modeling of a comonomer photopoly-merization system using high-throughput conversion data. Macromolecules 41:230–237

22. Brocchini S, James K, Tangpasuthadol V, Kohn J (1997) A combinatorial approach for polymerdesign. J Am Chem Soc 119:4553–4554

23. Akinc A, Anderson Daniel G, Lynn David M, Langer R (2003) Synthesis of poly(β -aminoester)s optimized for highly effective gene delivery. Bioconjug Chem 14:979–988

24. Anderson DG, Lynn DM, Langer R (2003) Semi-automated synthesis and screening of a largelibrary of degradable cationic polymers for gene delivery. Angew Chem, Int Ed 42:3153–3158

25. Becer CR, Paulus RM, Hoogenboom R, Schubert US (2006) Optimization of the nitroxide-mediated radical polymerization conditions for styrene and tert-butyl acrylate in an automatedparallel synthesizer. J Polym Sci Part A Polym Chem 44:6202–6213

26. Eggenhuisen TM, Becer CR, Fijten MWM, Eckardt R, Hoogenboom R, Schubert US (2008)Libraries of statistical hydroxypropyl acrylate containing copolymers with LCST propertiesprepared by NMP. Macromolecules 41:5132–5140

27. Fijten MWM, Meier MAR, Hoogenboom R, Schubert US (2004) Automated parallel inves-tigations/optimizations of the reversible addition-fragmentation chain transfer polymerizationof methyl methacrylate. J Polym Sci Part A Polym Chem 42:5775–5783

28. Fijten MWM, Paulus RM, Schubert US (2005) Systematic parallel investigation of RAFT poly-merizations for eight different (meth)acrylates: a basis for the designed synthesis of block andrandom copolymers. J Polym Sci Part A Polym Chem 43:3831–3839

29. Paulus RM, Fijten MWM, de la Mar MJ, Hoogenboom R, Schubert US (2005) Reversible add-ition-fragmentation chain transfer polymerization on different synthesizer platforms. QSARComb Sci 24:863–867

30. Zhang H, Abeln CH, Fijten MWM, Schubert US (2006) High-throughput experimentationapplied to atom-transfer radical polymerization: automated optimization of the copper catalystsremoval from polymers. e-Polymers

31. Zhang H, Fijten MWM, Hoogenboom R, Reinierkens R, Schubert US (2003) Application ofa parallel synthetic approach in atom-transfer radical polymerization: set-up and feasibilitydemonstration. Macromol Rapid Commun 24:81–86

32. Zhang H, Fijten MWM, Hoogenboom R, Schubert US (2003) Atom-transfer radical polymer-ization of methyl methacrylate utilizing an automated synthesizer. ACS Symp Ser 854:193–205

33. Zhang H, Marin V, Fijten MWM, Schubert US (2004) High-throughput experimentation inatom-transfer radical polymerization: a general approach toward a directed design and under-standing of optimal catalytic systems. J Polym Sci Part A Polym Chem 42:1876–1885

34. Adams N, Gans B-JD, Kozodaev D, Sanchez C, Bastiaansen CWM, Broer DJ, SchubertUS (2006) High-throughput screening and optimization of photoembossed relief structures.J Comb Chem 8:184–191

35. Hoogenboom R, Fijten MWM, Schubert US (2004) Parallel kinetic investigation of 2-oxazo-line polymerizations with different initiators as basis for designed copolymer synthesis.J Polym Sci Part A Polym Chem 42:1830–1840

36. Hoogenboom R, Fijten MWM, Schubert US (2004) The effect of temperature on the livingcationic polymerization of 2-phenyl-2-oxazoline explored utilizing an automated synthesizer.Macromol Rapid Commun 25:339–343


37. Hoogenboom R, Fijten MWM, Wijnans S, Van den Berg AMJ, Thijs HML, SchubertUS (2006) High-throughput synthesis and screening of a library of random and gradientcopoly(2-oxazoline)s. J Comb Chem 8:145–148

38. Hoogenboom R, Thijs HML, Fijten MWM, Schubert US (2007) Synthesis, characterization,and cross-linking of a library of statistical copolymers based on 2-“soy alkyl”-2-oxazoline and2-ethyl-2-oxazoline. J Polym Sci Part A Polym Chem 45:5371–5379

39. Guerrero-Sanchez C, Abeln C, Schubert US (2005) Automated parallel anionic polymeriza-tions: enhancing the possibilities of a widely used technique in polymer synthesis. J Polym SciPart A Polym Chem 43:4151–4160

40. Guerrero-Sanchez C, Schubert US (2004) Towards automated parallel anionic polymeriza-tions. Polymeric Mater Sci Eng 90:647–648

41. Becer CR, Hahn S, Fijten MWM, Thijs HML, Hoogenboom R, Schubert US (2008) Librariesof methacrylic acid and oligo(ethylene glycol) methacrylate copolymers with LCST behavior.J Polym Sci Part A Polym Chem 46:7138–7147

42. Fijten MWM, Kranenburg JM, Thijs HML, Paulus RM, Van Lankvelt BM, D Hullu J, Spring-intveld M, Thielen DJG, Tweedie CA, Hoogenboom R, VanVliet KJ, Schubert US (2007)Synthesis and structure–property relationships of random and block copolymers: a direct com-parison for copoly(2-oxazoline)s. Macromolecules 40:5879–5886

43. Hoeppener S, Wiesbrock F, Hoogenboom R, Thijs HML, Schubert US (2006) Morphologiesof spin-coated films of a library of diblock copoly(2-oxazoline)s and their correlation to thecorresponding surface energies. Macromol Rapid Commun 27:405–411

44. Meier MAR, Aerts SNH, Staal BBP, Rasa M, Schubert US (2005) PEO-b-PCL block copoly-mers: synthesis, detailed characterization, and selected micellar drug encapsulation behavior.Macromol Rapid Commun 26:1918–1924

45. Meier MAR, Gohy J-F, Fustin C-A, Schubert US (2004) Combinatorial synthesis of star-shaped block copolymers: host-guest chemistry of unimolecular reversed micelles. J Am ChemSoc 126:11517–11521

46. Meier MAR, Schubert US (2005) Combinatorial evaluation of the host-guest chemistry of star-shaped block copolymers. J Comb Chem 7:356–359

47. Lohmeijer BGG, Wouters D, Yin Z, Schubert US (2004) Block copolymer libraries usingsupramolecular strategies. Polym Mater Sci Eng 90:723–724

48. Schmatloch S, Van den Berg AMJ, Fijten MMW, Schubert US (2004) Automated parallelsynthesis of metallo-supramolecular polymers. Polym Mater Sci Eng 90:645–646

49. Schmatloch S, van den Berg AMJ, Fijten MWM, Schubert US (2004) A high-throughput ap-proach towards tailor-made water-soluble metallo-supramolecular polymers. Macromol RapidCommun 25:321–325

50. Ekin A, Webster DC (2006) Library synthesis and characterization of 3-aminopropyl-terminated poly(dimethylsiloxane)s and poly(e-caprolactone)-b-poly(dimethylsiloxane)s.J Polym Sci Part A Polym Chem 44:4880–4894

51. Ekin A, Webster DC. (2006) Synthesis and characterization of novel hydroxyalkyl carbamateand dihydroxyalkyl carbamate terminated poly(dimethylsiloxane) oligomers and their blockcopolymers with poly(e-caprolactone). Macromolecules 39:8659–8668

52. Pieper R, Ekin A, Webster DC, Casse F, Callow JA, Callow M, E. (2007) A combinatorialapproach to study the effect of acrylic polyol composition on the pProperties of crosslinkedsiloxane-polyurethane fouling-release coatings. J Coatings Techn Res 4:453–461

53. Webster DC, Bennett J, Kuebler S, Kossuth MB, Jonasdottir S (2004) High throughput work-flow for the development of coatings. J Coatings Tech 1:34–39

54. Nasrullah MJ, Webster DC (2006) Polymerization of styrene and t-butyl acrylate by atom-transfer radical polymerization – high throughput approach. Polymer Prepr 47:217–218

55. Nastrullah MJ, Ekin A, Bahr JA, Gallagher-Lein C, Webster DC (2006) Practical and auto-mated high throughput approach: atom-transfer radical polymerization of styrene and t-butylacrylate. PMSE Prepr 95:10–12

56. Rojas R, Harris NK, Piotrowska K, Kohn J (2009) Evaluation of automated synthesis for chainand step-growth polymerizations: can robots replace the chemists? J Poly Sci Part A PolymChem 47:48–58


57. Kamau SD, Hodge P, Williams RT, Stagnaro P, Conzatti L (2008) High throughput synthesis ofpolyesters using entropically-driven ring-opening polymerizations. J Comb Chem 10:644–654

58. Nasrullah MJ, Bahr JA, Gallagher-Lein C, Webster DC, Roesler RR, Schmitt P (2009) Au-tomated parallel polyurethane dispersion synthesis and characterization J Coatings Tech Res6:1–10

59. Hoogenboom R, Fijten MWM, Abeln CH, Schubert US (2004) High-throughput investigationof polymerization kinetics by online monitoring of GPC and GC. Macromol Rapid Commun25:237–242

60. Cawse JN, Olson D, Chisholm BJ, Brennan M, Sun T, Flanagan W, Akhave J, Mehrabi A,Saunders D (2003) Combinatorial chemistry methods for coating development V: generating acombinatorial array of uniform coatings samples. Prog Org Coatings 47:128–135

61. Chisholm B, Potyrailo R, Cawse J, Shaffer R, Brennan M, Molaison C, Whisenhunt D,Flanagan B, Olson D, Akhave J, Saunders D, Mehrabi A, Licon M (2002) The developmentof combinatorial chemistry methods for coating development I. Overview of the experimentalfactory. Prog Org Coatings 45:313–321

62. de Gans B-J, Duineveld PC, Schubert US (2004) Inkjet printing of polymers: state-of-the-artand future developments. Adv Mater 16:203–213

63. de Gans B-J, Kazancioglu E, Meyer W, Schubert US (2004) Ink-jet printing polymers andpolymer libraries using micropipettes. Macromol Rapid Commun 25:292–296

64. de Gans B-J, Schubert US (2003) Inkjet printing of polymer micro-arrays and libraries: instru-mentation, requirements, and perspectives. Macromol Rapid Commun 24:659–666

65. Tekin E, de Gans B-J, Schubert US (2004) Ink-jet printing of polymers – from single dots tothin film libraries. J Mater Chem 14:2627–2632

66. Anderson DG, Levenberg S, Langer R (2004) Nanoliter-scale synthesis of arrayed biomaterialsand application to human embryonic stem cells. Nat Biotech 22:863–866

67. Diaz-Mochon JJ, Bialy L, Keinicke L, Bradley M (2005) Combinatorial libraries – from solu-tion to 2D microarrays. Chem Commun 1384–1386

68. Tourniaire G, Collins J, Campbell S, Mizomoto H, Ogawa S, Thaburet J-F, Bradley M (2006)Polymer microarrays for cellular adhesion. Chemical Commun 2118–2120

69. Schmatloch S, Bach H, van Benthem RATM, Schubert US (2004) High-throughput experi-mentation in organic coating and thin film research: state-of-the-art and future perspectives.Macromol Rapid Commun 25:95–107

70. Iden R, Schrof W, Hadeler J, Lehmann S (2003) Combinatorial materials research in the poly-mer industry: speed versus flexibility. Macromol Rapid Commun 24:63–72

71. Majumdar P, Christianson DA, Roesler RR, Webster DC (2006) Optimization of coating filmdeposition when using an automated high throughput coating application unit. Prog Org Coat-ings 56:169–177

72. Smith JR, Kholodovych V, Knight D, Welsh WJ, Kohn J (2005) QSAR models for the analysisof bioresponse data from combinatorial libraries of biomaterials. QSAR Comb Sci 24:99–113

73. Weber N, Bolikal D, Bourke SL, Kohn J (2004) Small changes in the polymer structure in-fluence the adsorption behavior of fibrinogen on polymer surfaces: validation of a new rapidscreening technique. J Biomed Mater Res Part A 68A:496–503

74. Smith JR, Seyda A, Weber N, Knight D, Abramson S, Kohn J (2004) Integration of combi-natorial synthesis, rapid screening, and computational modeling in biomaterials development.Macromol Rapid Commun 25:127–140

75. Abramson SD, Alexe G, Hammer PL, Kohn J (2005) A computational approach to predictingcell growth on polymeric biomaterials. J Biomed Mater Res Part A 73A:116–124

76. Meier MAR, Schubert US (2006) Selected successful approaches in combinatorial materialsresearch. Soft Matter 2:371–376. This paragraph was partially reproduced by permission ofThe Royal Society of Chemistry:http://dx.doi.org/10.1039/b518304a

77. Vogel BM, Cabral JT, Eidelman N, Narasimhan B, Mallapragada SK (2005) Parallel synthesisand high throughput dissolution testing of biodegradable polyanhydride copolymers. J CombChem 7:921–928

78. Reynolds CH (1999) Designing diverse and focused combinatorial libraries of synthetic poly-mers. J Comb Chem 1:297–306


79. Kim D-Y, Dordick JS (2001) Combinatorial array-based enzymatic polyester synthesis.Biotechnol Bioeng 76:200–206

80. Lynn DM, Anderson DG, Putnam D, Langer R (2001) Accelerated discovery of synthetictransfection vectors: parallel synthesis and screening of a degradable polymer library. J AmChem Soc 123:8155–8156

81. Akinc A, Lynn DM, Anderson DG, Langer R (2003) Parallel synthesis and biophysical charac-terization of a degradable polymer library for gene delivery. J Am Chem Soc 125:5316–5323

82. Anderson DG, Peng W, Akinc A, Hossain N, Kohn A, Padera R, Langer R, Sawicki JA (2004)A polymer library approach to suicide gene therapy for cancer. Proc Natl Acad Sci U S A101:16028–16033

83. Lavastre O, Illitchev I, Jegou G, Dixneuf PH (2002) Discovery of new fluorescent materialsfrom fast synthesis and screening of conjugated polymers. J Am Chem Soc 124:5278–5279

84. Bosman AW, Heumann A, Klaerner G, Benoit D, Frechet JMJ, Hawker CJ (2001) High-throughput synthesis of nanoscale materials: structural optimization of fFunctionalized one-step star polymers. J Am Chem Soc 123:6461–6462

85. Connal LA, Vestberg R, Hawker CJ, Qiao GG (2007) Synthesis of dendron functionalized corecross-linked star polymers. Macromolecules 40:7855–7863

86. Malkoch M, Schleicher K, Drockenmuller E, Hawker CJ, Russell TP, Wu P, Fokin VV (2005)Structurally diverse dendritic libraries: a highly efficient functionalization approach usingclick-chemistry. Macromolecules 38:3663–3678

87. Percec V, Mitchell CM, Cho W-D, Uchida S, Glodde M, Ungar G, Zeng X, Liu Y,Balagurusamy VSK, Heiney PA (2004) Designing libraries of first generation AB3 and AB2self-assembling dendrons via the primary structure generated from combinations of (AB)y-AB3 and (AB)y-AB2 building blocks. J Am Chem Soc 126:6078–6094

88. Meier MAR, Filali M, Gohy J-F, Schubert US (2006) Star-shaped block copolymer stabilizedpalladium nanoparticles for efficient catalytic Heck cross-coupling reactions. J Mater Chem16:3001–3006

89. Rasa M, Meier MAR, Schubert US (2007) Transport of guest molecules by unimolecularmicelles evidenced in analytical ultracentrifugation experiments. Macromol Rapid Commun28:1429–1433

90. Webster DC, Chisholm BJ, Stafslien SJ (2007) Mini-review: combinatorial approaches for thedesign of novel coating systems. Biofouling 23:179–192

91. Potyrailo RA, Chisholm BJ, Olson DR, Brennan MJ, Molaison CA (2002) Development ofcombinatorial chemistry methods for coatings: high-throughput screening of abrasion resis-tance of coatings libraries. Anal Chem 74:5105–5111

92. Ribeiro E, Stafslien SJ, Casse F, Callow JA, Callow ME, Pieper RJ, Daniels JW, Bahr JA, Web-ster DC (2008) Automated image-based method for laboratory screening of coating librariesfor adhesion of algae and bacterial biofilms. J Comb Chem 10:586–594

93. Casse F, Ribeiro E, Ekin A, Webster DC, Callow JA, Callow ME (2007) Laboratory screeningof coating libraries for algal adhesion. Biofouling 23:267–276

94. Becer CR, Schubert US (2009) Parallel optimization and high-throughput preparation of well-defined copolymer libraries using controlled/“living” polymerization methods. Adv Polym SciDOI 10.1007/12_2009_16

95. Fasolka MJ, Stafford CM, Beers KL (2009) Gradient and microfluidic library approaches topolymer interfaces. Adv Polym Sci DOI 10.1007/12_2009_17



Parallel Optimization and High-ThroughputPreparation of Well-Defined CopolymerLibraries Using Controlled/“Living”Polymerization Methods

C. Remzi Becer and Ulrich S. Schubert

Abstract This chapter highlights the application of controlled/“living” polymer-ization (CLP) techniques in automated parallel synthesizers for both optimizingreaction parameters as well as preparing copolymer libraries. Special attention isgiven to the use of CLP techniques for constructing well-defined copolymer li-braries. Furthermore, alternative strategies for the preparation of block copolymerlibraries are discussed.

Keywords Automated parallel synthesis · Block copolymers · High-throughputexperimentation · Polymer libraries · Random copolymers

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192 Parallel Optimization of Controlled/“Living” Polymerizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.1 Radical Polymerization Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.2 Ionic Polymerization Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3 Synthesis of Well-Defined Copolymer Libraries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353.1 Preparation via Controlled Radical Polymerization Techniques . . . . . . . . . . . . . . . . . . . . 35

C.R. Becer and U.S. Schubert (�)Laboratory of Macromolecular Chemistry and Nanoscience, Eindhoven University of Technology,Den Dolech 2, 5612, AZ, Eindhoven, The NetherlandsandLaboratory of Organic and Macromolecular Chemistry, Friedrich-Schiller-University Jena,Humboldtstr., 10, 07743 Jena, GermanyandDutch Polymer Institute, John F. Kennedylaan 2, 5612, AB, Eindhoven, The Netherlandse-mail:[email protected];[email protected]

e-mail: [email protected]; [email protected]

18 C.R. Becer and U.S. Schubert

3.2 Preparation via Ionic Polymerization Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463.3 Supramolecular Synthesis – LEGO R© Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

Abbreviations

AA Acrylic acidAcBr Acetyl bromideAcCl Acetyl chlorideAcI Acetyl iodideAFM Atomic force microscopyAIBN α,α-AzobisisobutyronitrileAmor N-Acryoyl morpholineATRP Atom transfer radical polymerizationBEB (1-Bromo ethyl) benzenebpy 4,4′-Dialkyl substituted bipyridineBrEBiB 2-Bromo-2-methylpropanoyl bromideCBDB 2-Cyano-2-butyl dithio benzoateCLP Controlled/“living” polymerizationCROP Cationic ring opening polymerizationCRP Controlled radical polymerizationCTA Chain transfer agentDMA N,N-Dimethyl acrylamideDMAc N,N-Dimethyl acetamideDMAEMA N,N-Dimethyl aminoethyl acrylamideDMF N,N-Dimethyl formamideDP Degree of polymerizationDSC Differential scanning calorimetryEEA 1-Ethoxy ethyl acrylateEBIB Ethyl-2-bromo-iso-butyrateEtOx 2-Ethyl-2-oxazolineGC Gas chromatographyHPA 2-Hydoxypropyl acrylateiPrOx 2-iso-Propyl-2-oxazolineLCST Lower critical solution temperatureMA Methyl acrylateMAA Methacrylic acidMADIX Macromolecular design via the interchange of xanthatesMBP Methyl bromo propionateMMA Methyl methacrylateMeOMA 2-Methoxyethyl 2-methylacrylateMeO2MA 2-(2-Methoxyethoxy)ethyl 2-methylacrylate

Parallel Optimization and High-Throughput Preparation 19

MeOx 2-Methyl-2-oxazolineMn Number average molar massnBA n-Butyl acrylateNIPAM N-Isopropyl acrylamideNMP Nitroxide mediated polymerizationNMR Nuclear magnetic resonanceNonOx 2-Nonyl-2-oxazolineOEGMA Oligo(ethyleneglycol) methyl ether methacrylateOEGEMA Oligo(ethylene glycol) ethyl ether methacrylatePAA Poly(acrylic acid)PDI Polydispersity indexPEEA Poly(1-ethoxyethyl acrylate)PEG Poly(ethyleneglycol)PEO Poly(ethylene oxide)PheOx 2-Phenyl-2-oxazolinePDMAEMA Poly(N,N-dimethyl aminoethyl methacrylate)PMA Poly(methyl acrylate)PMMA Poly(methyl methacrylate)PnBA Poly(n-butyl acrylate)PSt Poly(styrene)PtBA Poly(tert-butyl acrylate)RAFT Reversible addition-fragmentation chain transferSEC Size exclusion chromatographys-BuLi sec-ButyllithiumSPE Solid phase extractionSoyOx 2-“Soyalkyl”-2-oxazolineSt StyrenetBA tert-Butyl acrylateTEMPO 2,2,6,6-Tetramethyl-1-piperidinyloxy stable radicalTg Glass transition temperatureTGA Thermal gravimetric analysisTsCl p-Toluene sulfonyl chloride

1 Introduction

Tailor-made macromolecules have come into the focus of polymer science toovercome the challenges of a number of complex applications from the nano to themacro scale. Materials scientists have been designing and synthesizing tailor-mademacromolecules specific for each application. These materials are composed of dif-ferent monomeric units, chemical functionalities, and topologies. The challenge hasbeen to control precisely the position of the functionality on the polymer, to deter-mine the necessary ratio of monomeric units, as well as to understand the effect ofthe molecular architecture on the material performance.


The development of controlled/“living” polymerization (CLP) techniques hasopened a new window to researchers to gain control successfully over the syn-thesis of well-defined polymeric structures [1–5]. However, each polymerizationtechnique requires specific catalysts, initiators, and optimum reaction conditionsfor different monomers. Optimization and understanding the effect of each inputvariable on the polymerization kinetics or the obtained macromolecule can be per-formed by the use of high-throughput experimentation techniques. There is nodoubt that screening a wide range of reaction parameters will allow researchers todefine the most efficient synthesis conditions for successfully designing and prepar-ing tailor-made polymers. Following the identification of the optimum reactionparameters, systematic sets of copolymers can be synthesized to elucidate structure-property relationships. Alternatively, gradient thin film libraries allow screening theeffect of two or more parameters on a relatively small scale [6–8]. There are severalmethods to create gradient thin film libraries, e.g., flow-coating and ink jet printing[9, 10]. Countless data sets are being obtained by analysis of large polymer librariesand these data sets allow pinpointing of the best performance materials [11–15].

In this chapter, we will focus on the use of CLP techniques for the synthe-sis of systematic copolymer libraries using high-throughput approaches. Priorto that, automated parallel optimization reactions that have been performed fordifferent CLP techniques will be discussed. At the end of this chapter there will bea highlight on the latest synthetic approaches to synthesize well-defined polymerlibraries.

2 Parallel Optimization of Controlled/“Living” Polymerizations

Absolute structural control over the polymer chain represents the primary target inmodern synthetic polymer chemistry. In practice, the term “well-defined” is com-monly used for polymers that exhibit low polydispersity indices; however theirstructural composition should also be known in some detail. Important key featurescan be listed such as the chemical structures of the initiating and the end groups,monomer composition, number of repeating units as well as topology. Different syn-thetic approaches have been developed to gain control over the architecture. Ionicand radical polymerization techniques have been the most promising ones to providethe desired macromolecules. There is still the need to develop and optimize thesetechniques further, not only to improve the synthesis procedures but also to pro-vide a deeper insight into the fundamentals of polymerization mechanisms.

Researchers have already invested several decades to elucidate the effect of inputvariables on the polymerization kinetics and the polymer structures. Many researchgroups have devoted their resources to obtaining reproducible data on polymeriza-tion kinetics. One of the methods to achieve that is to conduct several experiments inparallel to keep most reaction inputs constant and to minimize unpredictable envi-ronmental effects. In these series of experiments it appeared to be necessary to apply


automated parallel synthesis platforms and standardized experimental protocols inorder to provide extended and comparable data sets within a short period of time.

2.1 Radical Polymerization Techniques

Starting from 1956, living ionic polymerizations became the major interest for thesynthesis of well-defined polymers. Szwarc reported that in the anionic polymeriza-tion of styrene (St) the polymer chains grew until all the monomer was consumed;the chains continued to grow upon addition of more monomer [16].

According to the IUPAC definition, ionic polymerization is a type of chain poly-merization where the kinetic-chain carriers are ions or ion pairs [17]. However, thesetechniques have some limitations such as the necessity of extreme purity of thechemicals and the reaction medium, incompatibility between the reactive centersand monomers, and the sensitivity to certain chemical functionalities that limitsthe monomer selection. These challenges directed researchers to discover or de-velop alternative polymerization techniques. One of the alternative polymerizationroutes is radical polymerization since it is less discriminating regarding the typesof polymerizable vinyl monomers and more tolerant to several functionalities. Themost common method is free radical polymerization, which results in polymers withbroad molar mass distributions. However, polymers with relatively high polydisper-sity indices may be of some advantage in industrial processing. For instance, lowmolar mass polymer chains in polymers with broad molar mass distributions pro-vide a plasticizer effect during processing. However, these ill-defined polymers arenot suited for advanced applications and are also not suitable for understandingstructure-property relationships.

As a consequence of the free radical polymerization kinetics, the terminationrates are extremely fast in comparison to the slow initiation rates. This results in theformation of high molar mass chains at the initial stage of the polymerization anddecreasing molar masses in the latter stages due to the decrease in the monomerconcentration. Under these circumstances, broad molar mass distributions are in-evitable.

There were several attempts to gain better control on the free radicalpolymerization process [18, 19]. One of these methods was named the “iniferter”method. The compounds used in this technique can serve as initiator, transferagent and terminating agent [20–22]. Another technique is based on the use ofbulky organic compounds such as diaryl or triarylmethyl derivatives [23–25]. Themain disadvantages of these systems comprise slow initiation, slow exchange,direct reaction of counter radicals with monomers, and their thermal decomposi-tion. Therefore, these techniques did not offer the desired level of control over thepolymerization.

Relatively new controlled radical polymerization (CRP) methods, which werediscovered in the mid-1990s, focused on establishing a precise equilibrium betweenthe active and dormant species. Three approaches, namely atom transfer radical


polymerization (ATRP) [1, 2], nitroxide mediated polymerization (NMP) [3, 26],and reversible addition fragmentation chain transfer (RAFT) [4, 27], out of severalothers, have attracted the most attention due to their success in providing relativelystable chain end functionalities that can be reactivated for subsequent block copoly-merizations or post polymerization modifications.

2.1.1 Atom Transfer Radical Polymerization

ATRP has become the most widely applied CRP technique due to its simple mech-anism and commercially available reagents. This technique was first reported in1995, independently by Sawamoto and Matyjaszewski [28, 29]. The polymeriza-tion mechanism is based on the reversible redox reaction between alkyl halides andtransition metal complexes. Scheme 1 illustrates the mechanism of normal ATRP.

The simplicity of the polymerization reaction is the result of intense researchcarried out by several groups on the importance and the fundamentals of eachparameter. In particular, Matyjaszewski et al. have spent great effort on the construc-tion of numerous comparison charts on the activity of initiators and ligands that areused in ATRP [30–32]. These published comparison tables represent the summaryof hundreds of single experiments and are now a very important and reliable sourceof data for the ATRP technique.

It is obvious that automated parallel synthesis robots provide a great advantageto the researcher that has necessarily to keep all secondary parameters constantthroughout the screening reactions. These commercially available robotic systemshave been constantly tested not only by its producers but also by academic groupsto verify the reproducibility of the high-throughput experimentation setups. Re-cently, we have reported a standard protocol on the automated kinetic investigationof controlled/living radical polymerization of various monomers as a first step toobtain comparable results independent of the research group [33].

Automated parallel experiments were carried out to rapidly screen and optimizethe reaction conditions for ATRP of methyl methacrylate (MMA) [34]. A set of108 different reactions was designed for this purpose. Different initiators and differ-ent metal salts have been used, namely ethyl-2-bromo-iso-butyrate (EBIB), methylbromo propionate (MBP), (1-bromo ethyl) benzene (BEB), and p-toluene sulfonylchloride (TsCl), and CuBr, CuCl, CuSCN, FeBr2, and FeCl2, respectively. 2,2′-Bipyridine and its derivatives were used as ligands. The overall reaction schemeand the structure of the used reagents are shown in Scheme 2.

R-X + Mtn/Ligand

kact

kdeactR + Mt

n+1/Ligand

kp

R-R / RH & R=

l

+M

k t

Scheme 1 General mechanism of atom transfer radical polymerization (ATRP)


RX

O

O

p-Xylene, 90οC

Metal salt/Ligand

O

O

R

X

n

RX = EBIB, MBP, BEB, TsClMetal salt=CuBr, CuCl, CuSCN, FeBr2, FeCl2Ligand=bpy, dMbpy, dHbpy, dTbpy, 4,5'-dMbpy, 5,5'-dMbpy, 4Mbpy, 6Mbpy

O

O

Br H

O

O

BrBr

SO2Cl

(EBIB) (MBP) (BEB) (TsCl)

N N

R R'

R4 = R4' = H (bpy), CH3 (dMbpy), n-C6H13 (dHbpy), n-C9H19 (dNbpy), n-C13H27 (dTbpy)R4 = R5' = CH3 (4,5'-dMbpy), R5 = R5' = CH3 (5,5'-dMbpy)R4= CH3 and R' = H (4-Mbpy), R6 = CH3 and R' = H (6-Mbpy)

Scheme 2 Schematic representation of the ATRP of MMA using different initiators, ligands andmetal salts. [MMA]0 : [initiator]0 : [metal salt]0 : [ligand]0 = 150:1:1:2 and MMA : p-xylene =1:2 v/v

High-throughput experimentation of the ATRP of MMA was carried out in aChemspeed ASW2000 automated synthesizer to screen rapidly and to optimize thereaction conditions. Two reactor blocks were used in parallel and each block con-sisted of 16 reaction vessels equipped with a double jacket heater. The typical layoutof the automated synthesis platform is illustrated in Fig. 1. There are several loca-tions for the reactor blocks in the platform and most commonly one or two blocksare used in parallel in order to keep the high-throughput workflow running with-out any bottlenecks. The stock solution rack is equipped with an argon inlet tokeep the stock solutions under inert conditions. A solid phase extraction (SPE) unit,which is equipped with alumina columns, is used to remove the metal salt from thealiquots. The samples intended for characterization are transferred into small vials


SEC

Reactor blocks SPE unit Stock solutions

SEC samples GC samples M

M

M

M

N

N

N

N

T

T

T

T

H

H

H

H

CB

CC

CS

FB

M

M

M

M

N

N

N

N

T

T

T

T

H

H

H

H

M

M

M

N

N

N

N

T

T

T

T

H

H

H

H

M

M

M

M

N

N

N

N

T

T

T

H

H

H

M

M

M

M

N

N

N

N

T

T

T

T

H

H

H

H

EBIB

MBP

BEB

TsCl

EBIB

MBP

BEB

TsCl

CuBr CuCl CuSCN FeBr2 FeCl2bpy

FC

5

8

6

7

1

4

2

3

FC

FC

FC

CS

CS

CS

CB

CB

CB

CC

CC

CC

FB

FB

FB

Fig. 1 Schematic representation of the automated synthesizer and combinations of metal salts,initiators and ligands used in this study. The symbols used in this figure are as follows: dMbpy, M;dHbpy, N; dTbpy, T; CuBr, CB; CuCl, CC; CuSCN, CS; FeBr2, FB; FeCl2, FC; CuBr + ligand +TsCl (ligand = 4,5′ −dMbpy, 1; 5,5′ −dMbpy, 2; 4Mbpy, 3; and 6Mbpy, 4), and CuCl+ ligand+TsCl (ligand = 4,5′ − dMbpy, 5; 5,5′ − dMbpy, 6; 4Mbpy, 7; and 6Mbpy, 8). (Reprinted withpermission from [34]. Copyright (2004) John Wiley & Sons, Inc.)

arranged in racks and the racks are transferred to the autosampler of the analyticalinstruments, such as gas chromatography (GC) or gas chromatography coupled withmass spectrometry (GC–MS), or size exclusion chromatography (SEC). In addi-tion, there is an injection port for online SEC measurements. The technical detailsand further explanation on this system can be found in several reviews [35–40]. Itshould be noted that the computer-based planning and robotic performing of thereactions as well as the utilization of fast characterization techniques significantlydecreased the research time for the designed library from several months to twoweeks. The experimental results obtained could be compared and used for elucida-tion of structure-property relationships of monomer, initiator, and catalytic systemssince all the reactions were carried under the same conditions.

Three main parameters were used to evaluate the efficiency of the polymer-ization, namely monomer conversion (CMMA), initiation efficiency of the reaction( f = Mn,theo/Mn,SEC), and polydispersity index (PDI). These results are depictedin Fig. 2. It is obvious that the Cu(I)-catalyzed systems are more effective thanthe Fe(II)-catalyzed systems under the studied conditions. It was concluded that abipyridine based ligand with a critical length of the substituted alkyl group (e.g.,dHbpy) shows the best performance in Cu(I)-mediated systems. Besides, Cu(I)halide-mediated ATRP with 4,5′-Mbpy as the ligand and TsCl as the initiator wasbetter controlled than that with dMbpy as the ligand, and polymers with much lowerPDI values were obtained in the former case.

Another challenge in ATRP is to remove the catalyst prior to the analysis of thepolymers. In the case of automated sample withdrawing, this leads to the necessityof an automated purification system. For this purpose, an SPE unit was utilized


c

dTbpy

dNbpy

dHbpydMbpy

bpy1.01.21.41.61.8

2.0PDI

b

0.00.20.40.60.81.0

f

dTbpy

dNbpy

dHbpy

dMbpy

bpy

CuBr

a

CuC1

CuSCN

FeBr2

FeCl2

CuBr

CuC1

CuSCN

FeBr2

FeCl2

CuBr

CuC1

CuSCN

FeBr2

FeCl2

0204060

80

CMMA

(%)

dTbpydNbpy

dHbpy

dMbpy

bpy

Fig. 2 Effects of metal salts, ligands, and initiators on CMMAs (a), f (b), PDIs (c) of the polymersin the atom transfer radical polymerization (ATRP) of methyl methacrylate (MMA) in p-xylene at90◦C. [MMA]0:[initiator]0:[metal salt]0:[ligand]0 = 150:1:1:2, MMA/p−xylene = 1:2v/v. EBIB,MBP, BEB, and TsCl were used as initiator from right to left in each ligand column, respectively(Reprinted with permission from [34]. Copyright (2004) John Wiley & Sons, Inc.)

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0300 400 500 600

a b

700 800

Wavelength (nm)

Abs

orba

nce

Fig. 3 UV-vis spectra of unpurified (solid line) and purified (dashed line) polymers in acetonitrileat same concentrations. (Reprinted with permission from [42]. Copyright (2003) John Wiley &Sons, Inc.)

to purify the aliquots [41, 42]. Different column materials were investigated anddeactivated aluminum oxide gave the best results. As shown in Fig. 3, the absorptionband of the sample decreased significantly after the purification.


2.1.2 Nitroxide Mediated Polymerization

Nitroxide mediated polymerization is one of the most environmentally friendlyCRP techniques and has a relatively simple polymerization mechanism sincethere is no need for a catalyst [3]. Solomon, Rizzardo and Moad have demon-strated the reaction between 2,2,6,6-tetramethyl-1-piperidinyloxy stable radical(TEMPO) and vinyl monomers in the range of the free radical polymerizationtemperature (40–60 ◦C) [43]. Since then, two different NMP concepts have beendeveloped, namely the bimolecular and the unimolecular process, respectively.Georges et al. described the bimolecular process for the preparation of low PDIvalue polystyrene initiated by benzoylperoxide and mediated by TEMPO [44].Following that, unimolecular initiators have been developed that have the similarconcept of well-defined initiators in living anionic and cationic processes [45, 46].In unimolecular polymerizations, the initiator and the mediator are combined ina single molecule (e.g., alkoxyamines) that also simplifies the polymerizationkinetics. The investigation on stable free nitroxide compounds were started withTEMPO and extended to several different types of nitroxide-containing compounds[3], such as phosphonate derivatives [47] or arenes [48]. The use of alkoxyaminesallows the greatest degree of control over the final polymeric structure with well-defined functional end groups. A schematic representation of the NMP of Stinitiated by N-(2-methylpropyl)-N-(1-diethylphosphono-2,2-dimethylpropyl)-O-(2-carboxylprop-2-yl) hydroxylamine (Bloc BuilderTM) is illustrated in Scheme 3.Bloc BuilderTM is an efficient alkoxyamine for styrenics as well as acrylates andcurrently commercially available from Arkema.

A systematic investigation has been performed on the homopolymerization ofSt and tert-butyl acrylate (tBA) in an automated parallel synthesizer [49]. TheChemspeed Accelerator SLT106TM was used in order to screen the effect of thepolymerization temperature by the use of an individually heated reactor block.These blocks allow conducting up to 16 parallel reactions each at different tem-peratures, heated by electrical heating and controlled by an individual heat sensor inevery reactor. The determination of the optimum polymerization temperature for aspecific nitroxide compound plays a crucial role in the control of the polymer chaingrowth. Relatively high temperatures will cause an increase in the apparent radi-cal concentration which will lead to increased side reactions such as terminationby coupling or disproportionation. An example of this behavior is visible in Fig. 4.The apparent polymerization rates were increased by higher temperatures. Highermonomer conversions were obtained at shorter reaction periods; however, the PDIvalues of the polymers were increased above 1.4. The optimum polymerization tem-perature range using this type of alkoxyamine (Bloc BuilderTM) was reported to bein the range of 110–125 ◦C.

The most critical point of all CRP techniques is to gain absolute control over theactivation and deactivation of the reactive chain end. This can be simply controlledby altering the polymerization temperature or increasing the deactivator concen-tration. Thus, additional stable free-nitroxide compounds can be added to the


Scheme 3 Schematic representation of the nitroxide mediated polymerization (NMP) ofstyrene (St)

0 2 4 6 8 100

1

2

3130 °C125 °C120 °C115 °C110 °C100 °C

90 °C

ln([

M] 0

/[M

] t)

Reaction time (h)0 20 40 60 80 100

Conversion (%)

0

5000

10000

15000

20000

Mn (

g/m

ol)

theoretical130 °C125 °C120 °C115 °C110 °C100 °C90 °C

1.01.21.41.6 P

DI

Fig. 4 Semilogarithmic kinetic plot for the NMP of styrene (St) initiated by Bloc BuilderTM (left),Mn and PDI values vs conversion plot at different polymerization temperatures (right). (Reprintedwith permission from [49]. Copyright (2006) John Wiley & Sons, Inc.)

polymerization medium besides the alkoxyamine initiator. This effect has beeninvestigated in detail for different types of monomers as shown in Scheme 4 [50].

Based on the optimization reactions described, the polymerization temperaturewas kept constant at 110 ◦C for the NMP of N,N-dimethyl acrylamide (DMA),N-acryoyl morpholine (Amor), and 2-hydoxypropyl acrylate (HPA). However, it


Scheme 4 Schematic representation of the chemical structures of the monomers N,N-dimethylacrylamide (DMA), N-acryoylmorpholine, and 2-hydroxypropyl acrylate

was necessary to add an excess of free nitroxide to obtain better control over thepolydispersity of the polymers. On the other hand, the excess of free nitroxidedramatically decreased the apparent polymerization rate. Parallel optimization re-actions were conducted for the monomers depicted in Scheme 4 in order to identifythe right balance between the polymerization rate and sufficient control over thepolydispersity. Some representative results from that report are illustrated in Fig. 5.All the reactions were performed in N,N-dimethyl formamide (DMF) as solvent,at a polymerization temperature of 110◦C, a monomer concentration of 2 M, anda monomer to initiator ratio of 100:1. The free nitroxide (SG-1) ratio was variedfrom 0 up to 20% with respect to the amount of initiator (Bloc BuilderTM). Ap-parently, the polymerization rate of Amor was decreased by increasing amounts ofSG-1, as shown in Fig. 5a. However, the lowest PDI values were obtained with thehighest amount of SG-1 (20%) (Fig. 5b). Similar trends of the molar masses andthe PDI values were observed for the polymerization of DMA and HPA. Accordingto the kinetic results, the apparent rates of Amor, DMA, and HPA were found tobe 16.0× 10−4 L mol−1 s−1,7.7 × 10−4 L mol−1 s−1, and 4.3× 10−4 L mol−1 s−1,respectively. In the case of copolymerization of these monomers, there will bedifferent distributions of monomers throughout the polymeric chain depending ontheir reactivity. The optimization of key parameters and the determination of thepolymerization rate constants provide critical knowledge on the construction ofpolymer libraries.

2.1.3 Reversible Addition Fragmentation Chain Transfer

The first well established RAFT polymerization using thiocarbonylthio compoundswas reported by CSIRO in 1998 [51]. Subsequently, another group reported a similarmechanism using xanthate RAFT agent; they named this technique macromolecular


0

1

2

3

4

5

6

70% SG1

ln([

M] 0

/[M

] t)

Reaction time (h)

3% SG16% SG1

10% SG1

0 4 8 12 16

15% SG120% SG1

0 20 40 60 80 100

2000

4000

6000

8000

10000

Conversion (%)

0% SG1 3% SG16% SG1

10% SG115% SG120% SG1Mn, th

1.001.251.501.75 P

DI

0 20 40 60 80 100

0

2000

4000

6000

8000

10000

Mn (

g/m

ol)

Mn (

g/m

ol)

Mn (

g/m

ol)

Conversion (%)

1.001.251.501.75

0% SG15% SG1

10% SG115% SG120% SG1M

n, th

PD

I

0 20 40 60 80 100

0

5000

10000

15000

20000

Conversion (%)

1.001.251.501.75

0% SG15% SG1

10% SG115% SG120% SG1M

n,th

PD

I

Fig. 5 Semilogarithmic kinetic plot of Amor (top left), molar mass and PDI value vs conversionplots for Amor (top right), N,N-dimethyl acrylamide (DMA) (bottom left), and 2-hydoxypropylacrylate (HPA) (bottom right). (Reprinted with permission from [50]. Copyright (2008) AmericanChemical Society)

design via the interchange of xanthates (MADIX) [52, 53]. RAFT polymerizationhas several advantages over other CRP techniques. The most significant advantageis the compatibility of the technique with a wide range of monomers, such as St,acrylates, methacrylates, and derivatives. This large number of monomers providesthe opportunity of creating well-defined polymer libraries by the combination ofdifferent monomeric units. The mechanism of the RAFT polymerization comprisesa sequence of addition-fragmentation processes as shown in Scheme 5.

The initiation and radical–radical termination reactions occur as in conventionalfree radical polymerization. This is followed by the addition of the propagatingspecies (A) to the chain transfer agent (CTA), which leads to the formation of anintermediate species (B). Therefore, a new radical (D) can be released to form newpropagating chains (E). In step IV, rapid equilibrium between active propagatingradicals and the corresponding dormant species provides equal probability for allchains to grow and allows for the production of polymers with low PDI values.Termination reactions occur via combination or disproportionation (step V) to someextent, but can be largely eliminated by maintaining appropriate conditions that con-trol the apparent radical concentration.


Scheme 5 Schematic representation of the mechanism of the RAFT polymerization

Some of the most important critical points in RAFT polymerizations are therelative concentrations of the free radical initiator, the CTA, and the monomer,since these will establish the delicate balance between the dormant and activespecies. Acrylate and methacrylate derivatives can be successfully polymerized us-ing 2-cyano-2-butyl dithio benzoate (CBDB) as a CTA. However, the amount of freeradical initiator (α,α-azobisisobutyronitrile (AIBN) is used in general) compared toCTA determines the rate of control over the polymerization. Therefore, eight dif-ferent acrylates or methacrylates were polymerized with different ratios of CTA toAIBN [54]. The structures of the monomers and the design of the experiment areshown in Fig. 6.

A reactor block consisting of 16 reactors was divided into 4 zones with 4 differ-ent CTA to initiator ratios, and 4 different acrylates or methacrylates were used ineach set of experiments. The polymerization of tert-butyl methacrylate was repeatedfour times to demonstrate the reproducibility of the polymerization in an automatedparallel synthesizer. Structural analysis of the polymers revealed that there was lessthan 10% deviation in the number average molar mass (Mn) and the PDI values.

The polymerization of four different acrylates at four different CTA to initia-tor ratios are shown in Fig. 7, as a representative example. The increased ratioof CTA to AIBN resulted in improved PDI values; however, there is a decrease


oo

o

o

oo

oo

o

Used acrylates

Used methacrylatesRaft : In

i ratio = 1 : 0.05

Raft : In

i ratio = 1 : 0.10R

aft

: In

i rat

io =

1 :

1.0

0R

aft

: In

i rat

io =

1 :

0.2

5

oo o o o

N

o o

Fig. 6 Schematic representation of the design of experiment and the structures of the used(meth)acrylates. (Reprinted with permission from [54]. Copyright (2005) John Wiley & Sons, Inc.)

Fig. 7 Mn and PDI values vs CTA to α,α-azobisisobutyronitrile (AIBN) ratio plots for differentacrylates. (Reprinted with permission from [54]. Copyright (2005) John Wiley & Sons, Inc.)

observed in the Mn of the polymers. All polymerizations were conducted at 70◦Cfor 10 h. Due to the different initiator concentrations, the rate of polymerizationdiffers and a significant decrease occurs in the molar mass for a certain reactiontime. Nevertheless, this systematic study not only proved the reproducibility of theRAFT polymerization of several (meth)acrylates but also provided the optimum ra-tio of CTA to initiator to be used in further reactions.


2.2 Ionic Polymerization Techniques

2.2.1 Anionic Polymerization

Anionic polymerization represents a powerful technique for synthesizing polymerswith low PDI values, thus providing good control over the chain length. This methodleads to less side reactions than radical polymerizations. For instance, unlike inradical polymerization, there is no termination by the combination of two activechains. However, the mechanism is more sensitive to impurities and functionalgroups, and therefore applicable for only a limited class of monomers.

It still represents a great challenge to conduct anionic polymerizations in anautomated parallel synthesizer. Above all, the technique requires an intensive purifi-cation of the reagents and the polymerization medium in order to obtain well-definedpolymers. Therefore, a special procedure has been described for the inertization ofthe reactors [55]. It is called “chemical cleaning,” which is essentially rinsing all thereactors with sec-butyllithium (s-BuLi) prior to the reaction in order to eliminate allchemical impurities. This process can be performed in an automated manner. Due tothe extreme sensitivity of the polymerization technique to oxygen, moisture, and im-purities, detailed investigations on the inertization procedure and the reproducibilityof the experiments need to be conducted.

The inertization procedure applied for the Chemspeed ASW2000 robot wasstarted with flushing the hood with argon overnight while the reactors were heatedup to 140 ◦C and were exposed to six cycles of vacuum (25 min)-argon (5 min)to eliminate oxygen and moisture completely. Afterwards, s-BuLi in cyclohexanewas added to all reactors and vortexed for 1.5 h at room temperature and 30 minat 50 ◦C with fresh cyclohexane solutions. As a final step, the washing solutionwas aspirated from all the reactors and one more vacuum–argon cycle was ap-plied to finalize the inertization of the automated parallel synthesizer for the anionicpolymerization.

The reproducibility of the reaction was examined by performing the parallel an-ionic polymerization of St. The polymerizations were performed in cyclohexaneand initiated by s-BuLi. The obtained polymers were analyzed by SEC and the dif-ference between the results was less than 3%. This corresponds to less than 5%deviation after calculating the real concentration of the initiator in the reactors by adouble titration method [56].

Once the robotic system and procedure passed the optimization and reproducibil-ity tests for a certain type of reaction, the researcher has the chance to move on tothe most delightful part of a high-throughput experimentation workflow that is tofollow the reaction kinetics of the reaction by withdrawing several samples undercomparable conditions. The characterization of these samples allows the determi-nation of the apparent rate constants and activation energies in a very reproducibleway. As an example, the anionic polymerization of St in cyclohexane initiated bys-BuLi under different reaction conditions was investigated. Several samples werewithdrawn during the reaction into small vials which were prefilled with 25μL of


0

1

2

3

4

5

6

0 500 1000 1500 2000 2500 3000 3500 4000 4500

Time (s)

-LN

(1-X

) T = 10 °C

T = 20 °C

T = 30 °C

T = 50 °C

T = 60 °C

0

5000

10000

15000

20000

25000

30000

35000

40000

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Conversion

Mn

1

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2

PD

I

T = 20 °C

T = 40 °C

T = 50 °C

T = 60 °C

Mn (th)

[Monomer] = 1.02 M, s-BuLi = 20 µL (sol. 1.1263 M) [Monomer] = 1.02 M, s-BuLi = 20 µL (sol. 1.1263 M)

Fig. 8 Semilogarithmic kinetic plot of the anionic polymerization of St (left) and molar mass andPDI values vs conversion plot (right). (Reprinted with permission from [55]. Copyright (2005)John Wiley & Sons, Inc.)

Table 1 Values of the activation energy reported in theliterature for the propagation reaction of the anionicpolymerization of styrene (St) in different solvents

SolventActivation energy(kJ mol−1) Reference

Ethylbenzene 75 [57]Benzene 59.9 [58]Toluene 64.8 [59]Toluene 59.9 [60]Cyclohexane 63±2 [55]

methanol to quench the polymerization. The monomer conversion and molar massesof each sample were determined by GC and SEC measurements. Figure 8 illustratesa representative example of the results obtained from these reactions.

Based on these results, the activation energy of the anionic polymerization of Stin cyclohexane was determined as 63± 2kJ mol−1 [55]. The results obtained werecomparable to the literature results obtained with other solvents and are summarizedin Table 1.

2.2.2 Cationic Ring Opening Polymerization

The living cationic ring opening polymerization (CROP) of 2-oxazolines was firstreported in the 1960s [61, 62]. The polymerization can be initiated by an electrophilesuch as benzyl halides, acetyl halides, and tosylate or triflate derivatives. The typicalpolymerization mechanism for 2-alkyl-2-oxazoline initiated by methyl tosylate isshown in Scheme 6.

The alkyl group attached at the 2 position of the 2-oxazoline providesextraordinary possibilities for variations in the monomer structure and the proper-


O

O

N O

TsO- TsO-

Termination

H2OOInitiation+ +

O S

N

N

O

O

Propagation NN

n+

O

NOH

n

O

Scheme 6 Schematic representation of the CROP of 2-ethyl-2-oxazoline (EtOx) initiated by themethyl tosylate

ties. This monomer family is a good candidate for high-throughput experimentationand allows creating different copolymer libraries by a combination of 2-oxazolineswith different side groups. However, the typical required polymerization times forthis type of monomers were previously in the range of 10–20 h. Nevertheless, thereaction time for 2-ethyl-2-oxazoline (EtOx) in acetonitrile could be reduced from6 h under standard conditions (oil bath heating, reflux at 82 ◦C) to less than 1 min(at 200 ◦C) under microwave irradiation. Thus, a high-throughput experimentationworkflow could be applied for CROP of 2-oxazolines. Several reaction parame-ters, such as temperature, pressure, and solvent were investigated under microwaveirradiation and using automated parallel synthesizers [63–67].

The living CROP of 2-methyl, 2-ethyl, 2-nonyl, and 2-phenyl-2-oxazolines(PheOx) were investigated at different temperatures in the range 80–200◦C us-ing a single mode microwave synthesizer [68]. The reaction rates were enhancedby a factor of up to 400. The livingness of the polymerization over the wholerange of polymerization temperatures was examined by following the first-orderkinetics of the monomer consumption. The semilogarithmic kinetic plots for 2-methyl-2-oxazoline (MeOx), EtOx, 2-nonyl-2-oxazoline (NonOx), and PhOx areshown in Fig. 9. All reactions show a linear increase which is an indication of aliving polymerization. Besides, each sample was characterized by SEC, and a linearincrease in their Mn was observed.

The apparent rate constants for CROP of each monomer at each investigatedtemperature were calculated and the corresponding activation energy plots were ob-tained. These plots are shown in Fig. 10 and they exhibit good agreement betweentheoretical and experimental data. It was concluded that a temperature of 140 ◦Crepresents the optimum polymerization temperature since it leads to almost perfectagreement with the theoretical values [69].

There are numerous reports available on the optimization of reaction conditionsof 2-oxazolines. For instance, the effect of solvent, temperature, pressure, monomerto initiator ratio, and many other critical parameters have been investigated to obtainthe optimum conditions [64–68]. Besides these parameters, the initiator structurehas also a great effect on the polymerization. The investigation on different initiatorstructures provided the necessary kinetic parameters for the use of functional ini-tiators [69]. Heterofunctional initiators have been used in polymer science for thecombination of different types of monomers that can be polymerized with differentpolymerization techniques, such as ATRP and CROP [70–72].


6 180 °C160 °C140 °C120 °C100 °C

5

4

3

2

1

00 10 20 30

2-Methyl-2-oxazoline

40 50 60

In ([M

0]/[M

t])

6

5

4

3

2

1

00 10 20 30

2-Nonyl-2-oxazoline

40 50 60

In ([M

0]/[M

t])

t /min

t /min

5

4

3

2

1

0

2-Ethyl-2-oxazoline

100806040200 120

In ([M

0]/[M

t])

t /min

5

4

3

2

1

0

2-Phenyl-2-oxazoline

100806040200 120

In ([M

0]/[M

t])

t /min

180 °C160 °C140 °C120 °C100 °C

180 °C

180 °C200 °C

160 °C

160 °C

140 °C

140 °C

120 °C

120 °C

100 °C80 °C

100°C

Fig. 9 Semilogarithmic kinetic plots for different 2-oxazolines at various temperatures. (Reprintedwith permission from [68]. Copyright (2005) American Chemical Society)

For instance, the CROP of EtOx using four different acetyl halide type of ini-tiators showed that the rate of polymerization increases with the decreased basicityof the counter ion: acetyl iodide < acetyl bromide < acetyl chloride. The apparentrates of polymerization of EtOx with different initiators are listed in Table 2.

3 Synthesis of Well-Defined Copolymer Libraries

3.1 Preparation via Controlled Radical PolymerizationTechniques

Free radical polymerization remains the most versatile technique due to its com-patibility with a wide range of monomers, its compatibility with protic andaqueous media, and experimentally less demanding conditions. Development of


12-Methyl-2-oxazoline

2-Nonyl-2-oxazoline

2-Ethyl-2-oxazoline

2-Phenyl-2-oxazoline

R2 = 0.99979

R2 = 0.99756

R2 = 0.99822

R2 = 0.99968

0 0

1

−1

In k

p −2

−3

−4

−5

−1

−2

−3

−4

−5

−62.2 2.3

103 T−1 (103 K−1)

2.4 2.5 2.6 2.7 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9

1

0

−1

−2

10

−1−2−3−4−5−6−7

−3

−4

−52.40 2.45 2.50 2.55 2.60 2.65 2.70 2.1 2.2 2.3 2.4 2.5 2.6

In k

p

In k

pIn

kp

103 T−1 (103 K−1) 103 T−1 (103 K−1)

103 T−1 (103 K−1)

Fig. 10 Arrhenius plots for the polymerization of various 2-oxazolines in acetonitrile. (Reprintedwith permission from [68]. Copyright (2005) American Chemical Society)

Table 2 Polymerization rates (in 10−3 L mol−1 s−1) of CROP of EtOx with different initiatorsat various temperatures – acetyl chloride (AcCl), acetyl bromide (AcBr), acetyl iodide (AcI), and2-bromo-2-methylpropanoyl bromide (BrEBiB)

Initiator 80◦C 90◦C 100 ◦C 120◦C 140 ◦C 160 ◦C 180◦C 200 ◦C 220◦C

AcCl – – – – – 11.4 47.4 111.3 126.8AcBr – – 7.8±0.1 15±1 54±4 149±1 342±18 – –AcI 3.5±0.3 7.7±0.3 14.3±0.1 42±1 150±9 351±1 – – –BrEBiB – – 7.9 24.9 44.6 202 351 – –

controlled/“living” radical polymerization techniques combined the advantagesof “living” polymerizations and free radical polymerizations. Thus, tailor-mademacromolecules could be synthesized from a variety of monomeric units bearingdifferent functional side groups. These special side groups bring an additional prop-erty to the polymeric material and influence the whole material properties such assolubility, mechanical properties, thermal properties as well as optical properties.

The combination of different monomeric units at various ratios generates totallynew materials, in most cases with the expected properties and rarely with unex-pected, superior properties. Therefore, the investigation of series of polymers inparallel has a great importance not only to elucidate structure-property relationshipsbut also to be able to realize “magic” compositions.


3.1.1 Random Copolymer Libraries

Copolymers containing oligo(ethyleneglycol) methyl ether methacrylate (OEGMA)and methacrylic acid (MAA) were synthesized via RAFT polymerization in anautomated synthesizer [73]. OEGMA containing polymers exhibit a phase transitionbehavior upon changes in temperature. This thermoresponsive behavior is based onthe formation or breakage of hydrogen bonds between OEGMA units and watermolecules at a critical temperature which is also known as lower critical solutiontemperature (LCST). Polymers with LCST behavior show a sudden and reversiblechange from hydrophilic to hydrophobic behavior that makes them attractive for us-age as “smart” switchable materials in applications ranging from, e.g., drug deliverysystems, soft actuators or valves, coatings to textile materials.

Thermoresponsive polymers have in common hydrogen donor or acceptorgroups mostly present on their side chains. The most well-known and investi-gated structures can be listed as N-isopropyl acrylamide (NIPAM), N,N-dimethylaminoethyl acrylamide (DMAEMA), 2-iso-propyl-2-oxazoline (iPrOx), andoligo(ethylene glycol) methacrylates [74]. Monomers composed of relatively shortpoly(ethyleneglycol) (PEG) chains and radically polymerizable methacrylate moi-eties represent versatile building blocks for thermoresponsive materials. Some of theOEGMA based on monomers, i.e., 2-methoxyethyl 2-methylacrylate (MeOMA),Me2OMA, OEGEMA246, OEGMA475, and OEGMA1100), are commercially avail-able and their chemical structures are schematically depicted in Scheme 7.

The monomers in Scheme 7 show increased hydrophilicity with increased sidechain length, and thereby the cloud points are expected to be higher. A wide range ofLCST values could be achieved by combining OEGMA monomers having short and

Scheme 7 Schematic representation of the chemical structures of monomers which may exhibitthermoresponsive behavior


60

55

50

Hydrophobic moieties

o o o o o

o

Hydrated side-chains

o

o9

om

cl

n

(R2 = 0.996)

y = 27.99 + 1.04x

45

40

35

30

25

LCS

T (

°C)

0 5 10 15

average number of OEGMA units per chain

20 25 30 35

25°C

37°C

Fig. 11 Plots of the measured lower critical solution temperature (LCST) as a function ofthe theoretical average number of OEGMA475 units per chain for a series of P(MeO2MA-co-OEGMA475) copolymers of various composition. Hydrophobic and hydrophilic molecular regionson the copolymer are indicated in red and blue, respectively. (Reprinted with permission from [76].Copyright (2008) John Wiley & Sons, Inc.)

long PEG side chains. Lutz et al. have investigated a series of these combinationsto determine the right balance between hydrophilicity and hydrophobicity [75]. Asshown in Fig. 11, the combination of OEGMA475 and 2-(2-methoxyethoxy)ethyl2-methylacrylate (MeO2MA) building blocks at various ratios results in thermore-sponsive polymers with cloud points in the range 27–60 ◦C [76].

Further detailed studies on the LCST properties of OEGMA based polymers havebeen conducted by Schubert et al. Several homopolymer and copolymer libraries ofthe monomers listed in Scheme 7 have been synthesized in an automated parallelsynthesizer using the RAFT polymerization process [74]. Homopolymers with dif-ferent chain lengths were prepared to understand the effect of the chain length onthe LCST behavior of this class of monomers. As expected, polymers containinglonger PEG chains as side groups revealed relatively higher cloud point tempera-tures than the others. For instance, cloud points for P(MeO2MA), P(OEGMA246),and P(OEGMA475) were found to be 21.8, 21.6, and 89.8 ◦C, respectively. All thepolymers were prepared with a 100 to 1 monomer to initiator ratio and their corre-sponding cloud points were measured in a buffer solution of pH 7. As the number ofhydrogen bond forming functionalities increases in the macromolecule, the energyrequired to break these bonds also increases. Therefore, homopolymers with longerPEG side chains required higher temperatures in order to break the hydrogen bondsand, as a result, precipitate in the solution. However, homopolymers of P(MAA) andP(OEGMA1100) with up to 100 repeating units did not show LCST behavior due totheir highly hydrophilic structures.

Copolymers of MAA and OEGMA were prepared via RAFT polymerizationin ethanol. A systematic parallel synthesis was performed to obtain copolymerscontaining different ratios of two monomers. Therefore, a complete screeningin composition of P(MAA)-r-(OEGMA)n copolymers was elaborated from 0%OEGMAn to 100% OEGMAn. As representative examples, the Mn and PDI valuesof two libraries, namely P(MAA)-r-(OEGMA475) and P(MAA)-r-(OEGMA1100),


0 20 40 60 80 1000

4000

8000

12000

16000

20000

24000

OEGMA475 mole% in the copolymer

1.01.11.21.3

P(MAA)-r-(OEGMA475) P(MAA)-r-(OEGMA1100)

PD

I

0 20 40 60 80 1000

10000

20000

30000

40000

50000

OEGMA1100 mole% in the copolymer

1.01.11.21.3 P

DI

Mn

(g/m

ol)

Mn(

g/m

ol)

Fig. 12 Mn and PDI values data of copolymers synthesized with different comonomer feed ratios.(Reprinted with permission from [73]. Copyright (2008) John Wiley & Sons, Inc.)

0 20 40 60 80 1000

20

40

60

80

100P(MAA)-r-(OEGMA475)

P(MAA)-r-(OEGMA1100)

Cal

cula

ted

OE

GM

Ax

(mol

e%)

Theoretical OEGMAx (mole%)

Fig. 13 Theoretical composition vs calculated composition. (Reprinted with permission from[73]. Copyright (2008) John Wiley & Sons, Inc.)

are depicted in Fig. 12. The Mn,GPC values of the copolymers increased linearlywith increasing OEGMAn content whereas the PDI values remained below 1.3.

Depending on the length of the side chain of the OEGMAn monomers, theircorresponding reactivity ratios are expected to be different. This may result in slightdifferences in the copolymer composition. Therefore, it is necessary to quantify theamount of each repeating unit in the copolymer. Based on 1H NMR spectroscopymeasurements the content of MAA and OEGMAn could be determined. Thus, theamount of OEGMAn repeating units in P(MAA)-r-(OEGMA)n copolymers werefound to be slightly higher than the theoretical values, as depicted in Fig. 13.

The LCST properties of P(MAA)-r-(OEGMA475) and P(MAA)-r-(OEGMA1100)libraries were determined in buffer solutions with different pH values. As illus-trated in Fig. 14, the copolymer library of P(MAA)-r-(OEGMA475) revealed cloudpoints from 25 to 90 ◦C at pH values of 2 and 4, respectively. It was mentioned


0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

Clo

ud p

oint

(°C

)

OEGMAn mole%

P((MAA)-r-(OEGMA))475pH 2pH 4pH 7

P((MAA)-r-(OEGMA))1100pH 2pH 4pH 7

Fig. 14 Cloud points of P(MAA)-r-(OEGMA475) and P(MAA)-r-(OEGMA1100) copolymer li-braries at different pH values. (Reprinted with permission from [73]. Copyright (2008) John Wiley& Sons, Inc.)

previously that homopolymers of P(MAA) and P(OEGMA)1100 did not show anyLCST due to their highly hydrophilic structures. However, the copolymers of thesetwo monomers do exhibit cloud points at certain comonomer compositions. Besides,these polymers were found to be double-responsive to both temperature and pH val-ues. For instance, the copolymer P(MAA)0.9-r-(OEGMA1100)0.1 is soluble at pH4 whereas it is not soluble at pH 2 in water at 37 ◦C. The determination of these“magic” compositions is easily feasible by screening large libraries of polymers forthe best performance.

The phase transition is directly related to the hydrophilic/hydrophobic balancein a copolymer and controlling the polymer composition provides a highly effec-tive way of tuning the LCST. Another example of responsive polymer libraries wasbased on the combination of 2-hydroxypropylacrylate and DMA or N-acryloyl mor-pholine [50]. The nitroxide mediated copolymerization conditions were chosen onthe basis of the kinetic investigation of the homopolymerizations, as discussed inthis chapter (see, e.g., Sect. 2.1.2).

The copolymers obtained for the P(Amor)-stat-(HPA) library (Scheme 8) re-vealed relatively low PDI values in the range from 1.16 to 1.32 and increasingMn,GPC values with increasing HPA content, as listed in Table 3. The observedcopolymerization rates for both monomers decreased with increasing HPA contentdue to the slower HPA-SG1 dissociation and association kinetics. The copolymercompositions were calculated from the monomer conversions obtained by GC aswell as from 1H NMR spectroscopy of the precipitated polymers.

The thermal transition behavior within the P(Amor)-stat-(HPA) copolymer li-brary was investigated using differential scanning calorimetry (DSC). For all mem-bers of this library, single glass transition temperatures (Tg) were obtained, which isan indication of a good mixing of the two different monomers. The homopolymer


O O

OH

O N

Bloc BuilderTM

O O

OH

O N

stat

O O

Scheme 8 Schematic representation of the synthesis of statistical copolymers of 2-hydroxypropylacrylate (mixture of isomers) and N-acryloyl morpholine

Table 3 Copolymerization results of the p(Amor-stat-HPA) library

Conversionb

(%) Mnc

(g mol−1)

CompositionGCb (mol%)

CompositionNMRd (mol%)

Namea Amor/HPA PDIc Amor/HPA Amor/HPA

A100 70/0e 6,900 1.32 100/0 100/0A90H10 63/69 7,700 1.27 89/11 90/10A80H20 48/51 7,200 1.22 79/21 78/22A70H30 56/53 8,300 1.26 71/29 68/32A60H40 45/38 8,100 1.21 64/36 58/42A50H50 45/42 8,500 1.23 52/48 47/53A40H60 37/28 8,300 1.20 47/53 38/62A30H70 37/31 8,800 1.20 34/66 29/72A20H80 34/28 8,400 1.20 23/77 18/82A10H90 28/20 8,100 1.16 13/87 7/93H100 0/22 8,200 1.16 0/100 0/100aNames indicate the monomer feed: A50H50 = p(Amor50-stat-HPA50)

bCalculated by GC using monomer/DMF ratios

cOf precipitated polymer, determined by GPC in DMAc using p(MMA) calibration

d1H NMR spectra were recorded in CDCl3

eConversion calculated by 1H NMR spectroscopy

of P(Amor) exhibits a Tg of 146.5 ◦C whereas it was found as 21.7 ◦C for P(HPA).Their copolymers exhibited Tg values in between these temperatures, as listed inTable 4. The cloud points were determined by turbidimetry measurements in a par-allel turbidimetry instrument (Crystal16, Avantium Technologies). This instrumentmeasures the turbidity from the transmission of red light through the sample vialas a function of temperature. One of the alternative techniques to measure turbidityis UV-Vis spectroscopy which requires at least 1 h per sample and per cycle. How-ever, the Crystal16 turbidimeter is capable of measuring 16 samples in parallel andrepeating as many cycles as programmed. Thus, the turbidimetry property of thesamples could be determined in an accelerated manner. As can be seen in Table 4,the cloud points of the P(Amor)-stat-(HPA) copolymers could be tuned from 20 to90 ◦C by varying the comonomer composition.


Table 4 Thermal and lower critical solution temperature (LCST) propertiesfor the copolymers of the p(Amor-stat-HPA) library

Compositiona Cloud pointc Cloud pointc

Name Amor/HPA (wt%) Tgb (◦ C) 0.5 wt% (◦ C) 1.0 wt% (◦ C)

A100 100/0 146.5 Soluble SolubleA90H10 90/10 130.6 Soluble SolubleA80H20 80/20 106.4 Soluble SolubleA70H30 69/31 95.8 Soluble SolubleA60H40 60/40 84.0 Soluble 88.0A50H50 49/51 75.2 79.5 65.9A40H60 40/60 61.4 62.7 53.0A30H70 30/70 51.8 49.2 38.3A20H80 19/82 41.7 41.5 30.9A10H90 8/82 31.3 33.9 25.3H100d 0/100 21.7 26.7 21.4d

aCalculated from 1H NMR spectroscopy

bMid-temperature

c50% transmittance point in first heating curve

dp(HPA) synthesized with 15 h reaction time

O O O

OH

N

Bloc BuilderTM

O O O

OH

N

stat

Scheme 9 Schematic representation of the synthesis of statistical copolymers of 2-hydroxypropylacrylate (mixture of isomers) and DMA

The copolymers obtained for the P(DMA)-stat-(HPA) (Scheme 9) library re-vealed relatively low PDI values below 1.3 and increasing Mn,GPC values withincreasing HPA content, as listed in Table 5. It should be noted that a poly(methylmethacrylate) (PMMA) calibration was used for the calculation of the Mn,GPC valuesand this causes an overestimation for HPA containing polymers. The copolymercompositions were calculated from the 1H NMR spectra; however, this methodwas not suitable for reliable conversion determination since the DMA-CH3 groupsoverlap in the 1H NMR spectra not only with the HPA-OH group but also withbroad backbone signals, which obstruct any reliable integration. Therefore, elemen-tal analysis was used as an alternative method for the calculation of the molecularcomposition of the copolymers.

Similarly, the thermal transition behavior of the members of the P(DMA)-stat-(HPA) copolymer library was investigated using DSC. For all members of thislibrary, single Tg were obtained, which is an indication of a good mixing of the two


Table 5 Copolymerization results of the p(DMA-stat-HPA) library

Conversionb

(%)CompositionNMRd (mol%)

Composition EA(mol%)

Namea DMA/HPA Mnc (g mol−1) PDIc DMA/HPA DMA/HPA

D100 53/0 4,500 1.23 100/0 100/0D90H10 60/93 6,800 1.20 85/15 88/12D80H20 46/36 6,500 1.23 75/25 78/22D70H30 67/73 8,400 1.27 69/31 68/32D60H40 57/56 8,500 1.24 53/48 59/41D50H50 66/53 9,800 1.27 46/54 50/50D40H60 58/53 9,600 1.26 33/67 41/59D30H70 82/29 10,900 1.24 28/72 32/68D20H80 57/44 10,700 1.22 18/82 22/78D10H90 71/48 10,500 1.20 9/91 12/88H100 0/33 11,100 1.21 0/100 0/100aNames indicate monomer feed: D50H50 = p(DMA50-stat-HPA50)

bCalculated by GC using monomer/DMF ratios

cDetermined by GPC in DMAc using p(MMA) calibration

d1HNMR spectra recorded in CDCl3

Table 6 Thermal and LCST properties for the copolymers of the p(DMA-stat-HPA) library

Compositiona Cloud pointc Cloud pointc

Name DMA/HPA (wt%) Tgb (◦ C) 0.5 wt% (◦ C) 1.0 wt% (◦ C)

D100 100/0 111.4 Soluble SolubleD90H10 85/15 97.6 Soluble SolubleD80H20 73/27 87.7 Soluble SolubleD70H30 62/38 79.7 Soluble SolubleD60H40 52/48 63.4 Soluble SolubleD50H50 43/57 58.8 Soluble 82.9D40H60 34/66 51.6 71.6 62.3D30H70 26/74 44.6 55.8 48.7D20H80 18/82 36.0 46.7 38.6D10H90 10/90 30.5 35.3 28.5H100 0/100 21.7 26.7 21.4aCalculated from elemental analysis

bMid-temperature

c50% transmittance point in first heating curve

different monomer structures. The homopolymer of P(DMA) has a Tg of 111.4 ◦Cwhereas a Tg of 21.7 ◦C was measured for P(HPA). Besides, their copolymersexhibited Tg values in between these temperatures, as listed in Table 6. The Tg

of P(DMA)-stat-(HPA) copolymers show a positive deviation from the Fox equa-tion, which is an indication of the presence of some weak hydrogen bonding ofthe hydroxyl group of HPA with the amide group of DMA. The cloud points were


determined by turbidimetry in the Crystal16. The phase transition temperature ofP(DMA)-stat-(HPA) copolymers could be tuned from 20 to 90 ◦C by varying thecomonomer composition.

3.1.2 Block Copolymer Libraries

The synthesis of well-defined block copolymers has been a challenge for decades.Block copolymers consist of segments with different solubility typically resultingin phase separation [77, 78] and solution aggregation behavior [79–81]. The effortsto synthesize them have strongly accelerated the development of CLP techniques.Several catalysts, functional initiators, and CTAs have been investigated for differentclasses of monomers to synthesize well-defined block copolymers. RAFT polymer-ization represents one of the most versatile techniques that can be applied for a widerange of monomers not only in organic solvents but also in aqueous media.

Poly(acrylic acid) (PAA) is a water soluble polymer that has been used in var-ious applications. Direct synthesis of well-defined PAA-containing polymers hasbeen a challenge for CRP techniques because of the acid-containing monomer. Sofar, RAFT polymerization and NMP techniques could be successfully employed forunprotected acrylic acid (AA) [82, 83]. Even though it is possible to polymerizeAA directly, the applied polymerization solvents had to be polar, implying thatblock copolymers with a variety of apolar monomers cannot be synthesized in astraightforward manner. Therefore, the protected analogues of AA, e.g., tBA [84]and benzyl acrylate [85], are often used for the polymerization and deprotectedfollowing the polymerization [86]. Du Prez et al. investigated a new route towardnear-monodisperse PAA and derived block copolymer structures by the RAFT poly-merization of 1-ethoxy ethyl acrylate (EEA) [87].

The temperature optimization for the RAFT polymerization of EAA revealed anoptimum reaction temperature of 70◦C. Block copolymers with a poly(methyl acry-late) (PMA), a poly(n-butyl acrylate) (PnBA), a PMMA, or a poly(N,N-dimethylaminoethyl methacrylate) (PDMAEMA) first block and a poly(1-ethoxyethyl acry-late) (PEEA) second block were successfully synthesized in an automated synthe-sizer. The synthesis robot was employed for the preparation of 16 block copolymersconsisting of 25 units of the first block composed of PMA (exp. 1–4), PnBA (exp.5–8), PMMA (exp. 9–13), and PDMAEMA (exp. 13–16) and a second block ofPEEA consisting of 25, 50, 75, or 100 units, respectively. The first blocks werepolymerized for 3 h and a sample from each reaction was withdrawn for SEC anal-ysis. Subsequently, EAA was added and the reactions were continued for 12 h. Themolar masses and PDI values of the obtained block copolymers are shown in Fig. 15.

The composition of the resulting block copolymers was further characterizedby 1H NMR spectroscopy and the results are summarized in Table 7. The integralvalues of the aromatic resonances of the RAFT agent were applied to calculatethe number average degree of polymerization (DP) for the monomers present inthe block copolymers. Deprotection of the PEEA containing block copolymers wasperformed in CHCl3 by heating to 80 ◦C for 3 h. The solutions exhibited a cloudy


1 2 3 4

0

5000

10000

15000

20000m

olec

ular

wei

ght (

dalto

n) Mn,th MA-b-EEA Mn MA-b-EEA Mn,th MA Mn MA

experiment number5 6 7 8

Mn,th n-BA-b-EEA Mn n-BA-b-EEA Mn,th n-BA Mn n-BA


Mn,th MMA-b-EEA Mn MMA-b-EEA Mn,th MMA Mn MMA


PD

I Mn,th DMAEMA-b-EEA Mn DMAEMA-b-EEA Mn,th DMAEMA Mn DMAEMA

experiment number

1.0

1.2

1.4

Fig. 15 Number average molar masses (Mn,GPC) and PDI values obtained for the first blocks andfor the final copolymers of PMA, PnBA, PMMA, or PDMAEMA (25 units) with PEEA (25, 50,75, and 100 units for 100% conversion). All Mn,GPC values are calculated against PMMA stan-dards. SEC eluent CHCl3:NEt3:i-PrOH. (Reprinted with permission from [87]. Copyright (2005)American Chemical Society)

Table 7 Compositions of the synthesized copolymers as determined by 1H NMRspectroscopy

Exp Mon A DPA,th DPEEA,th fEEA,th DPA,NMR DPEEA,NMR fEEA,NMR

1 MA 25 25 0.5 19 9 0.322 MA 25 50 0.67 17 22 0.563 MA 25 75 0.75 18 35 0.664 MA 25 100 0.8 20 52 0.725 n-BA 25 25 0.5 18 20 0.536 n-BA 25 50 0.67 18 44 0.717 n-BA 25 75 0.75 19 73 0.798 n-BA 25 100 0.8 22 87 0.809 MMA 25 25 0.5 23 6 0.2110 MMA 25 50 0.67 23 12 0.3411 MMA 25 75 0.75 23 20 0.4712 MMA 25 100 0.8 23 32 0.5813 DMAEMA 25 25 0.5 22 8 0.2714 DMAEMA 25 50 0.67 22 20 0.4815 DMAEMA 25 75 0.75 22 35 0.6116 DMAEMA 25 100 0.8 22 51 0.70

appearance which is an indication of the PAA formation. 1H NMR spectroscopyrevealed 85–100% deprotection for selected copolymers.

NMP is as successful as RAFT polymerization for the construction of blockcopolymers. A small library of block copolymers comprised of poly(styrene) (PSt)and poly(tert-butyl acrylate) (PtBA) was designed and the schematic representationof the reaction is depicted in Scheme 10 [49]. Prior to the block copolymeriza-tion, the optimization reactions for the homopolymerization of St and t-BA wereperformed as discussed in this chapter (e.g., see Sect. 2.1.2). Based on these results,


Scheme 10 Schematic representation of the nitroxide mediated block copolymerization of P(St)-b-(t-BA)

Table 8 Block copolymerization of poly(styrene)-b-(tert-butyl acrylate) at different macroinitia-tor to monomer ratios. PSn(n) = degree of polymerization of the PS macroinitiator

RunPSn(n) Initiator/t-BA

Time(h)

Conversion(%)

Mn,theo

(g mol−1)Mn,GPC

(g mol−1)PDI(Mw/Mn)

1 50 1:50 20 65 13,600 6,700 1.172 50 1:100 20 18 7,500 6,600 1.123 50 1:150 20 40 10,300 14,800 1.334 78 1:50 20 10 9,400 8,900 1.105 78 1:100 14 15 10,200 9,900 1.096 78 1:150 20 23 11,100 11,100 1.157 120 1: 50 14 11 14,800 14,800 1.138 120 1:100 14 10 14,100 15,600 1.109 120 1:150 14 13 14,500 17,300 1.11

PS macroinitiators were prepared with chain lengths of 50, 78, and 120, respectively.These three different PS macroinitiators were reacted with different amount of t-BAto obtain a 3× 3 library of P(St)-b-(t-BA). The resulting block copolymers werecharacterized by SEC to determine the Mn,GPC and PDI values, which are listed inTable 8.

3.2 Preparation via Ionic Polymerization Techniques

Ionic polymerization techniques are very powerful for the construction of well-defined block copolymers having controlled architectures, microstructures andmolar masses, narrow molar mass distributions, and chemical and compositionalhomogeneity. Under the appropriate experimental conditions, anionic polymer-izations are associated with the absence of any spontaneous termination or chaintransfer reactions. One important limitation of ionic polymerizations is the demand-ing experimental conditions required to achieve a living polymerization systemand its applicability to a rather narrow range of monomers. However, recent de-velopments not only in polymerization kinetics and reagents but also in synthesismethods and instrumentation have allowed extending the utility of the method to abroader range of monomers.


3.2.1 Random Copolymer Libraries

A library of random copolymers comprised of MeOx, EtOx, and NonOx has beenestablished, and the properties of the members have been studied [88]. Systematiccopolymerization studies and corresponding structure-property investigations havebeen performed in detail by Schubert et al. For this purpose, nine copolymers weresynthesized with 0–100 mol% (steps of 12.5 mol%) of the second monomer, result-ing in 27 polymerizations for three different combinations of MeOx, EtOx, andNonOx. The monomer conversion was followed by GC measurements. As shown inFig. 16, the content of the second monomer increases linearly with increasing molefraction of the second monomer, whereas the content of the first monomer decreaseslinearly.

The resulting semilogarithmic kinetic plots for the 50 mol% copolymerizationsare depicted in Fig. 17. The linearity in these first-order plots indicates a constantconcentration of the living polymer chains as expected for a living polymerization.The plots also revealed a slightly higher reactivity of MeOx in comparison to EtOxand NonOx. To elucidate further the copolymer compositions, the reactivity ratioswere determined from the relation between fraction of monomer A in the monomerfeed ( f1) and the incorporated fraction of monomer A at both ∼20 and ∼60%monomer conversion. The corresponding reactivity ratios calculated for MeOx,EtOx, and NonOx using two different methods, namely the Mayo-Lewis terminalmodel and the extended Kelen–Tüdös, are listed in Table 9.

The synthesis of statistical copolymers consisting of EtOx and 2-“soyalkyl”-2-oxazoline (SoyOx) via a microwave assisted CROP procedure was reported bySchubert et al. [89]. The SoyOx monomer is based on soybean fatty acids andhas an average of 1.5 double bonds per monomer unit. The designed polymer

0.4

0.3

ratio

[M]/[

DM

Ac]

0.2

0.1

0.00.0 0.2 0.4 0.6

fNonOx fNonOx fEtOx0.8 1.0

0.5a b c

0.4

0.3

ratio

[M]/[

DM

Ac]

0.2

0.1

0.00.0 0.2 0.4 0.6 0.8 1.0

0.5

0.4

0.3

ratio

[M]/[

DM

Ac]

0.2

0.1

0.00.0 0.2 0.4 0.6

MeOx

0.8 1.0

0.5

EtOxMeOxNonOx

EtOxNonOx

O N O N+ O N O N+ O N + O N

Fig. 16 Ratios of monomer to solvent ([M]/[DMAc]) obtained by GC for t = 0 samples of thedifferent copolymerizations plotted against fraction of the second monomer (f): EtOx:NonOx(a), MeOx:NonOx (b), MeOx:EtOx (c). These graphs clearly demonstrate the gradual changein monomer composition within the investigated library. (Reprinted with permission from [88].Copyright (2006) American Chemical Society)


00

O O OO O ON N NN N N

initialfinal

initialfinal

initialfinal

stage of polymerization: stage of polymerization: stage of polymerization:

f1 f1 f1

F1

F1

F1

conv conv conv

time (min) time (min) time (min)

EtOX EtOXMeOX MeOX

NonOx NonOx

In([M0/[Mt]) In([M0/[Mt]) In([M0/[Mt])

In([M0]/[M

t])

In([M0]/[M

t])

In([M0]/[M

t])Con

vers

ion

(%)

Con

vers

ion

(%)

Con

vers

ion

(%)

0.00.0

0.2

0.2

0.4

0.4

0.6

0.6

0.8

0.8

1.0

1.0 0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.00.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

0

1

2

3

4

0

1

2

3

4

0

1

2

3

4

20

40

60

80

100

0

20

40

60

80

100

0

20

40

60

80

100

200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200

a ec

fdb

+ + +

Fig. 17 Top row: conversion (ln([M]0/[M]t) against time plots for 50 mol% copolymerizations(a, c, e). Bottom row: relationship between the monomer feed ( f1) and the actual monomer in-corporation (F1) at the initial (∼20% conversion) and final (>50% conversion) polymerizationstages (b, d, f). Both conversion and monomer incorporation are shown for EtOx:NonOx (a, b),MeOx:NonOx (c, d), and MeOx:EtOx (e, f) copolymerizations. (Reprinted with permission from[88]. Copyright (2006) American Chemical Society)

Table 9 Reactivity ratios determined for 2-oxazoline copolymerizations utilizing both the Mayo-Lewis terminal model (MLTD) and the extended Kelen–Tüdös (KT) method. Initial defines ∼20%conversion and final defines >50% conversion

M1: M2 Method Initial r1 Initial r2 Final r1 Final r2

EtOx:NonOx MLTM 1.2±0.2 0.7±0.1 0.97±0.01 0.99±0.01EtOx:NonOx KT 1.23±0.13 0.60±0.05 0.91±0.05 0.94±0.03MeOx:NonOx MLTM 1.8±0.3 0.3±0.1 1.26±0.05 0.66±0.03MeOx:NonOx KT 1.94±0.15 0.25±0.04 1.83±0.04 0.46±0.02MeOx:EtOx MLTM 1.52±0.1 0.54±0.03 1.18±0.04 0.65±0.02MeOx:EtOx KT 1.67±0.04 0.51±0.04 1.63±0.05 0.52±0.04

library consisted of a series of copolymers in which the monomer compositionwas systematically varied allowing the determination of structure-property relation-ships. The monomer structures and the polymerization mechanism are depicted inScheme 11.

The polymerization mixtures consisting of EtOx, SoyOx, methyl tosylate, andacetonitrile were automatically prepared utilizing the liquid handling system of


Scheme 11 Schematic representation of the polymerization mechanism as well as the EtOx and2-“soyalkyl”-2-oxazoline (SoyOx) monomer structures

Table 10 Structural characterization of the synthesized P(EtOx)-stat-(SoyOx) copolymers

∼40% conversion Full conversionEntry fSoyOx,th fSoyOx

a Mn,GPCb PDI fSoyOx

a Mn,GPCb PDI

1 0 – – – 0 5,500 1.112 0.05 – – – 0.04 6,250 1.163 0.10 0.06 4,700 1.12 0.08 6,900 1.154 0.15 – – – 0.13 8,000 1.155 0.20 0.20 5,300 1.19 0.18 8,500 1.196 0.25 – – – 0.22 9,450 1.167 0.30 0.29 6,550 1.16 0.28 11,100 1.128 0.40 0.41 7,050 1.17 0.38 11,300 1.199 0.50 0.51 7,250 1.17 0.48 12,100 1.2610 0.60 0.63 7,950 1.18 0.58 12,700 1.3111 0.70 0.76 7,550 1.31 0.70 14,400 1.2812 0.80 0.88 9,100 1.24 0.81 13,100 1.4313 0.90 0.85 7,500 1.38 0.91 13,000 1.5414 1.00 – – – 1.0 15,000 1.75aDetermined by 1H NMR spectroscopy

bMn values are given in Dalton

the ASW2000 synthesis robot. The total degree of polymerization was aimed for100 and compositions of EtOx and SoyOx were altered in steps of 10 mol%. Theprepared vials were placed in the autosampler of the microwave synthesizer andwere irradiated one by one for a predefined reaction time and temperature. Thestructural characterizations of the resulting polymers were performed by SEC aswell as 1H NMR spectroscopy measurements and are summarized in Table 10. Thereactivity ratios of EtOx and SoyOx were examined and it could be concluded thatboth monomers have slightly higher reactivity to itself than to the other monomer.The Mayo-Lewis terminal model with nonlinear least square fitting of the data re-vealed reactivity ratios of rEtOx = 1.4± 0.3 and rSoyOx = 1.7± 0.3. These values


0 20 40 60 80 1000

20

40

60

80

100 Tm EtOx-SoyOx Tm EtOx-SoyOx UV-cured Tg EtOx-SoyOx Tg EtOx-SoyOx Uv-cured

tem

per

atu

re (

°C)

wt.% SoyOx

Fig. 18 Thermal properties (Tg and Tm) of the P(EtOx)-stat-(SoyOx) copolymer before and afterUV-curing. (Reprinted with permission from [89]. Copyright (2007) John Wiley & Sons, Inc.)

indicate that two different monomers will be present almost in a random fashion,whereby small clusters of the same monomer may be present in the polymer chains.

The unsaturated side chain of the SoyOx repeating units could be used for cross-linking well-defined P(EtOx)-stat-(SoyOx) copolymers. Thus, the effect of cross-linking on the thermal properties of the polymers was investigated. The thermalproperties of the synthesized P(EtOx)-stat-(SoyOx) copolymers before and afterUV-curing are illustrated in Fig. 18.

3.2.2 Block Copolymer Libraries

A library of 4 chain extended homopolymers and 12 diblock copoly(2-oxazoline)swas prepared from 2-methyl, 2-ethyl, 2-nonyl, PheOx in a very short period of time[90]. The CROP was initiated by methyl tosylate and performed in acetonitrile at140 ◦C in a single-mode microwave synthesizer. A total number of 100(50 + 50)repeating units was incorporated into the respective polymer chains. The concen-tration of the solutions and predefined polymerization times for each monomer andcomonomer are summarized in Table 11.

The structural characterization of the resulting diblock copolymers was per-formed by means of SEC, 1H NMR spectroscopy, thermal gravimetric analysis(TGA), and DSC. In most of the cases the PDI values were found to be lower than1.3. However, the calculation of the molar masses of the diblock copolymers wasnot straightforward since the calibration standards (PEG, PS, and PMMA) used inSEC systems do not provide accurate data. Moreover, the folding behavior of thedifferent block copolymers significantly influence the hydrodynamic volume and,


Table 11 Reaction time for the preparation of the diblock copoly(2-oxazolines). In each cell, thecorresponding entries indicate the initial concentration of the first monomer (first line), and thereaction times for the polymerization of the first and the second monomer, respectively (secondline)

First Second monomermonomer MeOx EtOx NonOx PhOx

MeOx4 M 4 M 4 M 4 M400s+400s 400s+500s 400s+400s 400s+1,800s

EtOx4 M 4 M 4 M 4 M500s+400s 500s+500s 500s+400s 500s+1,800s

NonOx2 M 2 M 2 M 2 M800s+800s 800s+1,000s 800s+800s 800s+3,600s

PhOx3 M 3 M 3 M 3 M2,400s+600s 2,400s+800s 2,400s+600s 2,400s+2,400s

Table 12 Theoretical number average molar masses (Mthn ) and polydispersity indices for the for

chain extended and 12 diblock copoly(2-oxazolines). In each cell, the first and second entry for thePDI values results from the GPC measurements in different eluents, chloroform and N,N-dimethylformamide (DMF), respectively

First monomer Second monomerMeOx EtOx NonOx PheOx

MeOxMth

n = 8.5kDa Mthn = 9.2kDa Mth

n = 14.2kDa Mthn = 11.6kDa

PDI: –/1.16 PDI: –/1.17 PDI: –/– PDI: –/1.25

EtOxMth



PDI: –/1.18 PDI: 1.12/1.16 PDI: 1.15/– PDI: 1.27/1.19

NonOxMth



PDI: –/– PDI: 1.64/– PDI: 1.14/– PDI: 1.24/–

PhOxMth



PDI: –/1.18 PDI: 1.35/1.19 PDI: 1.28/– PDI: 1.27/1.16

consequently, the Mn. Theoretical Mn and PDI values of the synthesized diblockcopolymer library are summarized in Table 12.

The glass-transition temperatures and the corresponding specific heats were mea-sured three times for each sample in order to enable the calculation of the standarddeviations, which were in the range of ±3% or lower. Apparently, the kind of sub-stituent greatly influences the Tg values, and rigid substituents (phenyl or methyl) orflexible substituents (ethyl or nonyl) cause an increase or decrease in correspondingTg values, respectively. The measured Tg values are plotted in Fig. 19.

A library of 30 triblock copolymers was synthesized from 2-methyl, 2-ethyl,2-nonyl, and PheOx in a single mode microwave synthesizer [92]. The polymers ex-hibited narrow PDI values and showed slight deviations from the targeted monomerratio of 33:33:33. The design of the experiments is shown in Scheme 12.


NonEt

EtNon

EtEt

MeEt

EtMe

MeN

on

NonM

e

MeM

e

MePhe

PheMe

EtPhe

PheEt

PhePhe

0

20

40

60

80

100

120

T g (°

C)

Fig. 19 Glass-transition temperatures for the chain extended and the diblock copoly(2-oxazoline)s(the error bars represent the range of standard deviation). Non50Non50, Non50Phe50, andPhe50Non50 did not exhibit a glass transition temperature (Tg). The stars represent the literaturevalues for Me50Me50, Et50Et50, and Non50Non50, respectively. (Reprinted with permission from[90]. Copyright (2005) American Chemical Society)

analysis 1stblockanalysis diblock copolymer

analysis diblock copolymer

analysis triblock copolymer analysis triblock copolymer analysis triblock copolymer

analysis 1st block

Scheme 12 Schematic representation of the synthetic procedure that was applied for the prepara-tion of three triblock copolymers with the same first and second blocks

The designed set of 2-oxazoline monomers that was used for the synthesis ofthe triblock copolymers (MeOx, EtOx, PheOx, and NonOx) yielded polymers ofdifferent polarity [91]. P(MeOx) and P(EtOx) are hydrophilic, whereas P(PheOx)and P(NonOx) are hydrophobic. All possible combinations of these four differentmonomers would result in 64 different structures. However, all polymers that wouldhave two times the same block after each other were excluded since they do repre-sent diblock copolymers. Additionally, some structures, which have NonOx as thefirst block and EtOx or MeOx as the second block, were excluded due to extensiveside reactions. Consequently, 30 different triblock copolymers were synthesized,and they are listed in Table 13 with their corresponding structural characterization.

The Tg values of all investigated triblock copolymers are plotted in Fig. 20. Theincreasing trend in the measured values is due to the incorporation of monomerswith rigid substituent such as methyl or phenyl. Moreover, the Tg values did not


Table 13 Number of incorporated monomer units into the 30 triblock copoly(2-oxazoline)s re-sulting from combined 1H NMR spectroscopy analyses (top) of the model [A and AB (blockco) polymers] and final polymers as well as the measured number average molar masses(Mn,SEC/PDI; bottom). 1H NMR spectra were recorded in CDCl3 or CD2Cl2 (PhOx containingpolymers) and GPC analyses were performed using DMF (with 5 mM NH4PF6) as eluent. Mn,GPCwas calculated utilizing poly(methyl methacrylate) (PMMA) standards

3rd block1st–2nd block MeOx EtOx PhOx NonOx

MeOx-EtOxMeOx-PhOx 33:31:33 33:33:36 – 33:30:32

14.1 kDa/1.22 13.9 kDa/1.15 10.2 kDa/1.21MeOx-NonOx 33:28:33 33:30:37 33:29:29

–9.9 kDa/1.20 10.0 kDa/1.21 10.6 kDa/1.27

EtOx-MeOx–

33:33:33 33:29:27 33:34:3110.9 kDa/1.32 12.4 kDa/1.23 9.5 kDa/1.28

EtOx-PhOx 33:31:30 33:30:33–

33:30:36

16.2 kDa/1.20 15.3 kDa/1.24 11.4 kDa/1.22EtOx-NonOx 33:33:37 33:33:33 33:33:31

–10.1 kDa/1.27 9.9 kDa/1.22 11.3 kDa/1.25

PhOx-MeOx–

33:35:35 33:27:33 33:31:31

15.3 kDa/1.21 15.2 kDa/1.19 9.1 kDa/1.23PhOx-EtOx 33:35:34 – 33:42:33 33:38:38

17.8 kDa/1.32 19.1 kDa/1.28 14.1 kDa/1.21PhOx-NonOx 33:38:34 33:45:37 33:36:33

–9.7 kDa/1.21 8.8 kDa/1.21 11.6 kDa/1.22

NonOx-PhOx 33:23:27 33:26:24–

33:32:337.2 kDa/1.40 7.8 kDa/1.33 10.3 kDa/1.38a

aGPC measurement with CHCl3:NEt3:2-PrOH (94:4:2) as eluent (PS calibration)

depend on the order of the blocks. It should also be noted that none of the triblockcopolymers showed more than one Tg value, indicating that there was no macro-scopic phase separation occurring in the bulk state. This is most likely due to therelatively short segments (33 repeating units) that were incorporated.

3.3 Supramolecular Synthesis – LEGO R© Approach

An alternative route to prepare well-defined block copolymers is first to prepare thehomopolymers with functional groups and then to connect them by noncovalent in-teractions [92–99]. A systematic 4×4 library of block copolymers based on PSt andPEG connected by an asymmetrical octahedral bis(terpyridine) ruthenium complexat the block junction was reported [78]. Moreover, the thin film morphology of thislibrary was investigated.


MeE

tNo

MeN

oEt

EtM

eNo

EtN

oMe

EtN

oEt

EtM

eEt Et

MeN

oMe

MeE

tMe

EtP

hEt

MeE

tPh

EtM

ePh

EtP

hMe

Me

PhM

eEt

PhE

tMe

MeP

hEt

MeP

hMe

PhE

tPh

PhM

ePh

Ph

0

20

40

60

80

100

Tg

(°C

)

Fig. 20 Glass transition temperatures of the triblock copoly(2-oxazoline)s and the P(MeOx),P(EtOx), and P(PheOx) homopolymers, sorted in ascending order. The polymers that contained(at least) one block of P(NonOx) and P(PheOx) at the same time did not exhibit any Tg in dif-ferential scanning calorimetry (DSC). (Me = P(MeOx),Et = P(EtOx),Non = P(NonOx), andPh = P(PheOx)). (Reprinted with permission from [91]. Copyright (2006) American ChemicalSociety)

Functional homopolymers can be synthesized by essentially two differentmethods. The first and more preferred way is to use a functional initiator which willensure a high rate of chain end functionality. For instance, the polymerization of Stinitiated by a unimolecular terpyridine-functionalized nitroxide initiator yields well-defined PS homopolymers. The second technique is based on post-polymerizationmodifications. In this case, the reaction between mPEG and chloroterpyridine yieldsterpyridine-functionalized PEG building blocks, as illustrated in Scheme 13.

The theoretical molar masses and the corresponding volume fractions of PS,the metal complex, and PEG content of the block copolymers are summarized inTable 14. The metal complex has been treated as the third block. All block copoly-mers have been purified by preparative SEC and column chromatography, withisolated yields between 10% and 80%. The expected ratios (within 10% error)for all components in the library were obtained from the integration of 1H NMRspectrum.

The morphology of this supramolecular diblock copolymer library has beeninvestigated by means of atomic force microscopy (AFM) measurements. As illus-trated in Fig. 21, at first glance different morphologies were obtained for differentcompositions. However, interpreting the phase behavior of supramolecular blockcopolymers is not straightforward. There are several important parameters thatplay a role in the phase behavior. For instance, the amorphous phase of PEG, thecrystalline phase of PEG, the metal complex, and the amorphous PSt contribute to


NN

N

O NN

N

Cl

ON

bulk125 oC

NN

N

O

ON

m

m

OH20

t-BuOKTHFreflux

20

N

N

N

O

OO

OH

n

NN

N

Cl

OO n

N

N

N

O

KOHDMSO70 oC

RuCl3DMF130 oC

OO n

N

N

N

O Ru Cl

Cl

Cl

N

N

N

O Ru

N

N

N

O PEOnPSm

PSm-[ PEOn-[

1) CHCl3:MeOH (4:1)N-ethylmorpholineref lux

2) NH4PF6

m = 20, 70, 200, 240 n = 70, 125, 225, 375

2+

Scheme 13 Schematic representation of the synthetic route towards a library of PStm-[Ru]-PEOn

block copolymers, where m and n denote the degree of polymerization (DP) of PSt and PEO,respectively, and where -[Ru]- represents the bis(terpyridine) ruthenium complex

the phase contrast. Besides, the final morphology is greatly affected by competitionsbetween self-organization, crystallization of the PEG block and vitrification of thePSt block [100].

4 Conclusion

Automated parallel synthesizers provide high-quality experimental data in rela-tively short periods of time. High-throughput experimentation techniques havebecome an inevitable reality in the field of polymer science, since there is a large


Fig. 21 Atomic force microscopy (AFM) phase images of all block copolymers in the libraryafter spin coating from 2% w/v solution in toluene. No annealing has been performed. The scalebar represents 100 nm. (Reprinted with permission from [78]. Copyright (2005) Royal Society ofChemistry)

parameter space not only including the reaction parameters but also the use ofdifferent monomers, catalysts, and polymerization techniques. The application ofCLP techniques in automated synthesizers have been demonstrated by several re-search groups. These techniques enable the synthesis of well-defined homo, block,or random copolymers and even more complex architectures such as graft, star, ordendritic shaped polymers.

The combination of CLP techniques and high-throughput experimentation toolsaccelerates the research in this field significantly. Besides, on the data collected, theconstruction of 3D-plots and extensive databases will provide the basis for deeperunderstanding of the underlying principles. As a consequence, the elucidation ofquantitative structure-property relationships will be feasible.


Table 14 The block copolymers in the library are displayed in the table by name, by theoreticalmolar masses and by the volume fractions of PSt, -[Ru]-, and poly(ethylene oxide) (PEO) (anno-tated between brackets)

PS20-[ PS70-[ PS200-[ PS240-[

PEO70-[RuCl3

PS20-[Ru]-PEO70

PS70-[Ru]-PEO70

PS200-[Ru]-PEO70

PS240-[Ru]-PEO70

Mn = 6,100 Da Mn = 11,400 Da Mn = 25,100 Da Mn = 29,300 Da(35:16:49) 1 (65:8:27) 2 (84:4:12) 3 (87:3:10) 4

PEO125-[RuCl3

PS20-[Ru]-PEO125

PS70-[Ru]-PEO125

PS200-[Ru]-PEO125

PS240-[Ru]-PEO125


PEO225-[RuCl3

PS20-[Ru]-PEO225

PS70-[Ru]-PEO225

PS200-[Ru]-PEO225

PS240-[Ru]-PEO225


PEO375-[RuCl3

PS20-[Ru]-PEO375

PS70-[Ru]-PEO375

PS200-[Ru]-PEO375

PS240-[Ru]-PEO375


Acknowledgement Financial support from the Dutch Polymer Institute (DPI project #502) isgratefully acknowledged.

References

1. Kamigaito M, Ando T, Sawamoto M (2001) Metal-catalyzed living radical polymerization.Chem Rev 101:3689–3745

2. Matyjaszewski K, Xia JH (2001) Atom transfer radical polymerization. Chem Rev 101:2921–2990

3. Hawker CJ, Bosman AW, Harth E (2001) New polymer synthesis by nitroxide mediated livingradical polymerizations. Chem Rev 101:3661–3688

4. Moad G, Rizzardo E, Thang SH (2005) Living radical polymerization by the RAFT process.Aust J Chem 58:379–410

5. Hadjichristidis N, Pitsikalis M, Pispas S et al. (2001) Polymers with complex architecture byliving anionic polymerization. Chem Rev 101:3747–3792

6. de Gans BJ, Duineveld P, Schubert US (2004) Ink-jet printing of polymers: state of the artand future developments. Adv Mater 16:203–213

7. Tekin E, Smith PJ, Schubert US (2008) Inkjet printing of functional materials: from polymersto nanoparticles and molecules. Soft Matter 4:703–713

8. de Gans BJ, Schubert US (2003) Ink-jet printing of polymer microarrays and libraries:requirements, possibilities and available instrumentation. Macromol Rapid Commun 24:659–666

9. Meredith JC, Smith AP, Karim A et al. (2000) Combinatorial materials science for polymerthin-film dewetting. Macromolecules 33:9747–9756

10. Smith AP, Douglas JF, Meredith JC et al. (2001) High-throughput characterization of pat-tern formation in symmetric diblock copolymer films. J Polym Sci Part B Polym Phys 39:2141–2158


11. Webster DC (2008) Combinatorial and high-throughput methods in macromolecular materi-als research and development. Macromol Chem Phys 209:237–246

12. Green JJ, Langer R, Anderson DG (2008) A combinatorial polymer library approach yieldsinsight into nonviral gene delivery. Acc Chem Res 41:749–759

13. Lynn DM, Anderson DG, Putnam D et al. (2001) Accelerated discovery of synthetic trans-fection vectors: parallel synthesis and screening of a degradable polymer library. J Am ChemSoc 123:8155–8156


15. Anderson DG, Peng W, Akinc A, Hossain N, Kohn A, Padera R, Langer R, Sawicki JA(2004) A polymer library approach to suicide gene therapy for cancer. Proc Nat Acad SciUSA 101:16028–16033

16. Szwarc M (1956) Living polymers. Nature 178:1168–116917. IUPAC (1997) Compendium of chemical terminology, (the “Gold Book”), 2nd edn. In:

McNaught AD, Wilkinson A (eds) Blackwell, Oxford (XML on-line corrected version:http://goldbook.iupac.org (2006-) created by M. Nic, J. Jirat, B. Kosata; updates compiledby A. Jenkins. ISBN 0–9678550–9–8)

18. Moad G, Solomon DH (1995) The chemistry of free radical polymerization. Elsevier Science,Bath

19. Sawamoto M, Kamigaito M (1999) In: Schlueter D (ed) Synthesis of polymers. VCH,Weinheim

20. Otsu T, Matsumoto A (1998) Controlled synthesis of polymers using the iniferter technique:developments in living radical polymerization. Adv Polym Sci 136:75–137

21. Ajayaghosh A, Francis R (1998) Narrow polydispersed reactive polymers by a photoiniti-ated free radical polymerization approach. Controlled polymerization of methyl methacrylate.Macromolecules 31:1436–1438

22. Ajayaghosh A, Francis R (1999) A xanthate-derived photoinitiator that recognizes and con-trols the free radical polymerization pathways of methyl methacrylate and styrene. J AmChem Soc 121:6599–6606

23. Borsig E, Lazar M, Capla M (1967) Polymerization of methyl methacrylate initiated by3,3,4,4-tetraphenyl hexane and 1,1,2,2-tetraphenyl cyclopentane. Makromol Chem 105:212

24. Sebenik A (1998) Living free-radical block copolymerization using thio-iniferters. ProgPolym Sci 23:875–917

25. Qin SH, Qiu KY, Swift G et al. (1999) “Living” radical polymerization of methyl methacry-late with diethyl 2,3-dicyano-2,3-diphenylsuccinate as a thermal iniferter. J Polym Sci Part APolym Chem 37:4610–4615

26. Moad G, Rizzardo E (1995) Alkoxyamine-initiated living radical polymerization: factors af-fecting alkoxyamine homolysis rates. Macromolecules 28:8772–8728

27. Moad G, Rizzardo E, Thang SH (2008) Toward living radical polymerization. Acc Chem Res41:1133–1142

28. Wang JS, Matyjaszewski K (1995) Controlled living radical polymerization – atom trans-fer radical polymerization in the presence of transition metal complexes. J Am Chem Soc117:5614–5615

29. Kato M, Kamigaito M, Sawamoto M et al. (1995) Polymerization of methyl methacry-late with the carbon-tetrachloride dichlorotris(triphenylphosphine)-ruthenium(II) methyla-luminum bis(2,6-di-tert-butylphenoxide) initiating system – possibility of living radicalpolymerization. Macromolecules 28:1721–1723

30. Kwak Y, Matyjaszewski K (2008) Effect of initiator and ligand structures on ATRP of styreneand methyl methacrylate initiated by alkyl dithiocarbamate. Macromolecules 41:6627–6635

31. Tang W, Matyjaszewski K (2007) Effects of initiator structure on activation rate constants inATRP. Macromolecules 40:1858–1863

32. Tang W, Kwak Y, Braunecker W et al. (2008) Understanding atom transfer radical polymer-ization: effect of ligand and initiator structures on the equilibrium constants. J Am Chem Soc130:10702–10713


33. Becer CR, Groth AM, Paulus RM et al. (2008) Protocol for automated kinetic investi-gation/optimization of the RAFT polymerization of various monomers. QSAR Comb Sci27:977–983

34. Zhang HQ, Marin V, Fijten MWM et al. (2004) High-throughput experimentation in ATRP:a general approach toward a directed design and understanding of optimal catalytic systems.J Polym Sci Part A Polym Chem 42:1876–1885

35. Meier MAR, Schubert US (2006) Selected successful approaches in combinatorial materialsresearch. Soft Matter 2:371–376

36. Meier MAR, Hoogenboom R, Schubert US (2004) Combinatorial methods, automated syn-thesis and high-throughput screening in polymer research: the evolution continues. MacromolRapid Commun 25:21–33

37. Hoogenboom R, Meier MAR, Schubert US (2003) Combinatorial methods, automated syn-thesis and high-throughput screening in polymer research: past and present. Macromol RapidCommun 24:15–32

38. Meier MAR, Schubert US (2004) Combinatorial polymer research and high-throughput ex-perimentation: powerful tools for the discovery and evaluation of new materials. J MaterChem 14:3289–3299

39. Zhang HQ, Hoogenboom R, Meier MAR et al. (2005) Combinatorial and high-throughputapproaches in polymer science. Meas Sci Technol 16:203–211

40. Guerrero-Sanchez G, Paulus RM, Fijten MWM et al. (2006) High-throughput experimen-tation in synthetic polymer chemistry: from RAFT and anionic polymerizations to processdevelopment. Appl Surf Sci 252:2555–2561

41. Zhang HQ, Abeln CH, Fijten MWM et al. (2006) High-throughput experimentation applied toatom transfer radical polymerization: automated optimization of the copper catalysts removalfrom polymers. e-polymers 8:1–9

42. Zhang HQ, Fijten MWM, Hoogenboom R et al. (2003) Application of a parallel syn-thetic approach in atom transfer radical polymerization: set up and feasibility demonstration.Macromol Rapid Commun 24:81–86

43. Moad G, Rizzardo E, Solomon DH (1982) Selectivity of the reaction of free radicals withstyrene. Macromolecules 15:909–914

44. Georges MK, Veregin RPN, Kazmaier PM et al. (1993) Narrow molecular weight resins by afree radical polymerization process. Macromolecules 26:2987–2988

45. Hawker CJ (1994) Molecular weight control by a living free radical process. J Am Chem Soc116:11185–11186

46. Hawker CJ, Barclay GG, Orellana A et al. (1996) Initiating systems for nitroxide-mediated “living” free radical polymerizations: synthesis and evaluation. Macromolecules29:5245–5254

47. Benoit D, Grimaldi S, Robin S et al. (2000) Kinetics and mechanism of controlled free-radical polymerization of styrene and n-butyl acrylate in the presence of an acyclic beta-phosphonylated nitroxide. J Am Chem Soc 122:5929–5939

48. Benoit D, Chaplinski V, Braslau R et al. (1999) Development of a universal alkoxyamine for“living” free radical polymerizations. J Am Chem Soc 121:3904–3920

49. Becer CR, Paulus RM, Hoogenboom R et al. (2006) Optimization of the NMRP conditionsfor styrene and tert-butylacrylate in an automated parallel synthesizer. J Polym Sci Part APolym Chem 44:6202–6213

50. Eggenhuisen TM, Becer CR, Fijten MWM et al. (2008) Libraries of statistical hydroxypropylacrylate containing copolymers with LCST properties prepared by NMP. Macromolecules41:5132–5140

51. Chiefari J, Chong YK, Ercole F (1998) Living free radical polymerization by reversibleaddition-fragmentation chain transfer – the RAFT process. Macromolecules 31:5559–5562

52. Corpart P, Charmot D, Biadatti T et al. (1999) Block polymer synthesis by controlled radicalpolymerization. (WO9858974) Chem Abstr 130:82018

53. Chapon P, Mignaud C, Lizarraga G et al. (2003) Automated parallel synthesis of MADIX(co)polymers. Macromol Rapid Commun 24:87–91


54. Fijten MWM, Paulus RM, Schubert US (2005) Systematic parallel investigation of RAFTpolymerizations for eight different (meth)acrylates: a basis for the designed synthesis of blockand random copolymers. J Polym Sci Part A Polym Chem 43:3831–3839

55. Guerrero-Sanchez C, Abeln C, Schubert US (2005) Automated parallel anionic polymeriza-tions: enhancing the possibilities of a widely used technique in polymer synthesis. J PolymSci Part A Polym Chem 43:4151–4160

56. Gilman H, Jaubein AH (1941) The quantitative analysis of alkyllithium compounds. J AmChem Soc 66:1515–1516

57. Hadjichristidis N, Iatrou H, Pispas S et al. (2000) Anionic polymerization: high vacuum tech-niques. J Polym Sci Part A Polym Chem 38:3211–3234

58. Glusker DL, Lysloff I, Stiles E (1961) Mechanism of anionic polymerization of methylmethacrylate II. Use of molecular weight distributions to establish a mechanism. J PolymSci 49:315–334

59. Auguste S, Edwards HGM, Johnson AF et al. (1996) Anionic polymerization of styreneand butadiene initiated by n-butyllithium in ethylbenzene: determination of the propaga-tion rate constants using Raman spectroscopy and gel permeation chromatography. Polymer37:3665–3673

60. Wang GM, van Beylen M (2003) Influence of π-complexing agents on the anionic polymer-ization of styrene with lithium as counterion in cyclohexane. 1. Effect of durene. Polymer44:6205–6210

61. Tomalia DA, Sheetz DP (1966) Homopolymerization of 2-alkyl and 2-aryl-2-oxazolines.J Polym Sci Part A Polym Chem 4:2253–2265

62. Seelinger W, Aufderhaar E, Diepers W et al. (1966) Recent synthesis and reactions of cyclicimidic esters. Angew Chem 20:913–927

63. Hoogenboom R, Fijten MWM, Paulus RM et al. (2006) Accelerated pressure synthesis andcharacterization of 2-oxazoline block copolymers. Polymer 47:75–84

64. Hoogenboom R, Fijten MWM, Schubert US (2004) The effect of temperature on the livingcationic polymerization of 2-phenyl-2-oxazoline explored utilizing an automated synthesizer.Macromol Rapid Commun 25:339–343

65. Hoogenboom R, Fijten MWM, Schubert US (2004) Parallel kinetic investigation of2-oxazoline polymerizations with different initiators as basis for designed copolymer syn-thesis. J Polym Sci Part A Polym Chem 42:1830–1840

66. Hoogenboom R, Wiesbrock F, Leenen MAM et al. (2005) Accelerating the living polymeriza-tion of 2-nonyl-2-oxazoline by implementing a microwave synthesizer into a high-throughputexperimentation workflow. J Comb Chem 7:10–13

67. Hoogenboom R, Paulus RM, Fijten MWM et al. (2005) Concentration effects in the CROPof 2-ethyl-2-oxazoline in N,N-dimethyl acetamide. J Polym Sci Part A Polym Chem 43:1487–1497

68. Wiesbrock F, Hoogenboom R, Leenen MAM et al. (2005) Investigation of the living cationicring-opening polymerization of 2-methyl, 2-ethyl, 2-nonyl, and 2-phenyl-2-oxazoline in asingle-mode microwave reactor. Macromolecules 38:5025–5034

69. Paulus RM, Becer CR, Hoogenboom R et al. (2008) Acetyl halide initiator screening forthe cationic ring opening polymerization of 2-ethyl-2-oxazoline. Macromol Chem Phys 209:794–800

70. Yagci Y, Tasdelen MA (2006) Mechanistic transformations involving living and con-trolled/living polymerization methods. Prog Polym Sci 31:1133–1170

71. Bernaerts KV, Du Prez FE (2006) Dual/heterofunctional initiators for the combination ofmechanistically distinct polymerization techniques. Prog Polym Sci 31:671–722

72. Becer CR, Paulus RM, Hoppener S et al. (2008) Synthesis of poly(2-ethyl-2-oxazoline)-b-poly(styrene) copolymers via a dual initiator route combining cationic ring openingpolymerization and atom transfer radical polymerization. Macromolecules 41:5210–5215

73. Becer CR, Hahn S, Fijten MWM et al. (2008) Libraries of MAA and OEGMA copolymerswith LCST behavior. J Polym Sci Part A Polym Chem 46:7138–7147


74. Fournier D, Hoogenboom R, Thijs HML, Paulus RM, Schubert US (2007) Tunable pH-and temperature-sensitive copolymer libraries by RAFT of methacrylates. Macromolecules40:915–920

75. Lutz JF, Hoth A (2006) Preparation of ideal PEG analogues with a tunable thermosensi-tivity by controlled radical copolymerization of 2-(2-methoxyethoxy)ethyl methacrylate andoligo(ethylene glycol) methacrylate. Macromolecules 39:893–896

76. Lutz JF (2008) Polymerization of oligo(ethylene glycol) (meth)acrylates: toward new gener-ations of smart biocompatible materials. J Polym Sci Part A Polym Chem 46:3459–3470

77. Thomas EL, Anderson DM, Henkee CS et al. (1988) Periodic area-minimizing surfaces inblock copolymers. Nature 334:598–601

78. Lohmeijer BGG, Wouters D, Yin ZH et al. (2004) Block copolymer libraries: modular versa-tility of the macromolecular Lego (R) system. Chem Commun 24:2886–2887

79. Pochan DJ, Chen Z, Cui H et al. (2004) Toroidal triblock copolymer assemblies. Science306:94–97

80. Jain S, Bates FS (2003) On the origins of morphological complexity in block copolymersurfactants. Science 300:460–464

81. Gohy JF (2005) Block copolymer micelles. Adv Polym Sci 190:65–13682. Ladaviere C, Dorr N, Claverie JP (2001) Controlled radical polymerization of acrylic acid in

protic media. Macromolecules 34:5370–537283. Couvreur L, Lefay C, Belleney J (2003) First nitroxide-mediated controlled free-radical poly-

merization of acrylic acid. Macromolecules 36:8260–826784. Haddleton DM, Crossman MC, Dana BH et al. (1999) Atom transfer polymerization of

methyl methacrylate mediated by alkylpyridylmethan-imine type ligands, copper(I) bromide,and alkyl halides in hydrocarbon solution. Macromolecules 32:2110–2119

85. Butun V, Vamvakaki M, Billingham NC et al. (2000) Synthesis and aqueous solution proper-ties of novel neutral/acidic block copolymers. Polymer 41:3173–3182

86. Mori H, Muller AHE (2003) New polymeric architectures with (meth)acrylic acid segments.Prog Polym Sci 28:1403–1439

87. Hoogenboom R, Schubert US, van Camp W et al. (2005) RAFT polymerization of 1-ethoxyethyl acrylate: a novel route toward near-monodisperse poly(acrylic acid) and derived blockcopolymer structures. Macromolecules 38:7653–7659

88. Hoogenboom R, Fijten MWM, Wijnans S et al. (2006) High-throughput synthesis and screen-ing of a library of random and gradient copoly(2-oxazoline)s. J Comb Chem 8:145–148

89. Hoogenboom R, Fijten MWM, Thijs HML et al. (2007) Synthesis, characterization, andcross-linking of a library of statistical copolymers based on 2-“soy alkyl”-2-oxazoline and2-ethyl-2-oxazoline. J Polym Sci Part:A Polym Chem 45:5371–5379

90. Wiesbrock F, Hoogenboom R, Leenen M et al. (2005) Microwave-assisted synthesis of a4× 2-membered library of diblock copoly(2-oxazoline)s and chain-extended homo poly(2-oxazoline)s and their thermal characterization. Macromolecules 38:7957–7966

91. Hoogenboom R, Wiesbrock F, Huang H et al. (2006) Microwave-assisted cationic ring-opening polymerization of 2-oxazolines: a powerful method for the synthesis of amphiphilictriblock copolymers. Macromolecules 39:4719–4725

92. Lohmeijer BGG, Schubert US (2002) Supramolecular engineering with macromolecules: analternative concept for block copolymers. Angew Chem Int Ed 41:3825–3829

93. Lohmeijer BGG, Schubert US (2005) The LEGO toolbox: supramolecular building blocks byNMP. J Polym Sci Part A Polym Chem 43:6331–6344

94. Lohmeijer BGG, Schubert US (2003) Water-soluble building blocks for metallo-supramo-lecular polymers: synthesis, complexation and decomplexation studies of poly(ethylene-oxide) moities. Macromol Chem Phys 204:1072–1078

95. Gohy JF, Lohmeijer BGG, Schubert US (2003) From supramolecular block copolymers toadvanced nano-objects. Chem Eur J 9:3472–3479

96. Meier MAR, Lohmeijer BGG, Schubert US (2003) Characterization of defined metal con-taining supramolecular block copolymers. Macromol Rapid Commun 24:852–857


97. Lohmeijer BGG, Schubert US (2004) Expanding the supramolecular polymer LEGO sys-tem: nitroxide mediated living free radical polymerization for metallo-supramolecular blockcopolymers with a polystyrene block. J Polym Sci Part A Polym Chem 42:4016–4027

98. Gohy JF, Lohmeijer BGG, Alexeev A, Wang XS, Manners I, Winnik MA, Schubert US (2004)Cylindrical micelles from the aqueous self-assembly of an amphiphilic poly(ethylene oxide)-b-poly(ferrocenylsilane) (PEO-b-PFS) block copolymer with a metallo-supramolecular linkerat the block junction. Chem Eur J 20:4315–4323

99. Hofmeier H, El-ghayoury A, Schenning APH, Schubert US (2004) New supramolecular poly-mers containing both terpyridine metal complexes and quadruple hydrogen bonding units.Chem Commun 318–319

100. Zhu L, Chen Y, Zhang AQ et al. (1999) Phase structures and morphologies determinedby competitions among self-organization, crystallization, and vitrification in a disor-dered poly(ethylene oxide)-b-polystyrene diblock copolymer. Phys Rev B Condens Matter60:10022–10031



Gradient and Microfluidic Library Approachesto Polymer Interfaces∗

Michael J. Fasolka, Christopher M. Stafford, and Kathryn L. Beers

Abstract We present an overview of research conducted at the National Instituteof Standards and Technology aimed at developing and applying combinatorial andhigh-throughput measurement approaches to polymer surfaces, interfaces and thinfilms. Topics include (1) the generation of continuous gradient techniques for fabri-cating combinatorial libraries of film thickness, temperature, surface chemistry andpolymer blend composition, (2) high-throughput measurement techniques for as-sessing the mechanical properties and adhesion of surfaces, interfaces and films, and(3) microfluidic approaches to synthesizing and analyzing libraries of interfacially-active polymer species.

Contents

1 Introduction: Surfaces and Interfaces in Polymer Scienceand Engineering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

2 Continuous Gradient Library Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 653 Gradient Library Fabrication Methods and Application Examples . . . . . . . . . . . . . . . . . . . . . . . . 66

3.1 Flow Coating: Polymer Film Thickness Gradients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 663.2 Gradient Hot Stage: Temperature Processing Libraries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 683.3 Surface Energy and Surface Chemistry Libraries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 703.4 Gradient Polymer Brush Libraries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 763.5 Polymer Blend Composition Gradients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

M.J. Fasolka (�), C.M. Stafford, and K.L. BeersMaterials Science and Engineering Laboratory, National Institute of Standards and Technology,Gaithersburg, MD 20899, USAe-mail: [email protected]; [email protected]; [email protected]

∗ Official contribution of the National Institute of Standards and Technology; not subject to copy-right in the United States. Certain commercial materials and equipment are identified in order tospecify adequately experimental procedures. In no case does such identification imply recommen-dation or endorsement by the National Institute of Standards and Technology, nor does it implythat the items identified are necessarily the best available for the purpose.

64 M.J. Fasolka et al.

4 High-Throughput Materials Testing: Surfaces, Interfaces, and Thin Films . . . . . . . . . . . . . . . . 844.1 Thin Film Mechanical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 844.2 Adhesion Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

5 High-Throughput Materials Synthesis and Solution Characterization:Microscale Approaches to PolymerLibrary Fabrication in Fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 945.1 Controlled Polymer Synthesis in Microchannels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 955.2 Characterization of Interfacially-Active Polymers in Microchannels. . . . . . . . . . . . . . . 96

6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

1 Introduction: Surfaces and Interfaces in Polymer Scienceand Engineering

The success of a huge range of polymer-based technologies, including advancedcoatings and adhesives, electronics materials, complex fluid formulations and bio-materials, hinges on the ability to produce tailored polymer surfaces and interfaces.This is because surface and interfacial properties govern key aspects of productstructure and performance, such as film and multilayer stability, mechanical reliabil-ity, adhesion, expression of functional moieties, component dispersion, and domainorientation, among others. Research dedicated to the understanding and engineeringof these factors is extensive, and has proceeded for a number of decades; this is dueto the fact that both the origins and effects of surface and interfacial properties arecomplex, depend upon a large number of variables, and can be difficult to predict.Both the importance and complexity of surface and interfacial science and engineer-ing make them excellent targets for combinatorial and high-throughput approaches.Indeed, some of the first uses of these methods for polymeric materials systems fo-cused on the formulation and performance testing of coatings [1], the behavior ofwhich depend greatly on surface and interfacial effects.

Starting in the late 1990s, and continuing for the following 10 years, the NationalInstitute of Standards and Technology (NIST) built and executed a research pro-gram that developed combinatorial methods aimed largely at addressing scientificand engineering challenges in polymer surfaces, interfaces and thin films. The NISTprogram, organized through the NIST Combinatorial Methods Center (NCMC,www.nist.gov/combi), concentrated on meeting two measurement-related needs inestablishing combinatorial approaches for polymer surfaces and interfaces: the de-sign and implementation of appropriate library fabrication and synthesis methods,and the development of high-throughput testing techniques to asses these libraries.This review surveys the research conducted through the NCMC with a focus onmethodology, technique development and descriptions of supporting case studies inpolymer surfaces, interfaces and films. The paper will start with a discussion of con-tinuous gradient techniques, where NIST was a pioneer in polymer materials. Thefabrication of gradient libraries for surfaces and interfaces will be considered next,including techniques for making continuous spreads in film thickness and composi-tion, surface chemistry and surface energy and temperature. Application studies willinclude film stability and wetting, polymer self-assembly, polymer brush behavior

Gradient and Microfluidic Library Approaches to Polymer Interfaces 65

and measurements, biomaterials surface engineering. The next section will considerNIST-developed high-throughput approaches to measuring surface, interface andfilm performance properties. These include rapid measurement of film modulus, ad-hesion, and interfacial strength. Applications examples include mechanical testingof ultrathin film systems, ultrasoft polymers, engineering adhesives and relativelyweak adhesive interactions. The final section will consider microfluidic and continu-ous microreactor approaches to polymer library fabrication and the high-throughputmeasurement of such systems. The primary focus will be on methods to produce sys-tematic libraries of interfacially-active polymer species, such as block copolymersand macromolecular surfactants. Measurement applications will include microflu-idic assessments of complex fluid structure, in particular solution self-assembly, andof fluid mixture properties such as interfacial tension.

2 Continuous Gradient Library Techniques

A key challenge in combinatorial research is the creation of specimen libraries thatexhibit a diversity of composition, processing conditions and other parameters overprescribed ranges. A common design for achieving this is the so-called “discrete”library, which consists of a large collection of individual sub-specimens. The mainadvantage of the discrete approach is the ability to incorporate a great number of dif-ferent parameters in a single library – in this sense the design is versatile. However,since the library parameter space is divided into discrete sub-specimens, each witha single set of parameters, it is possible to “skip over” what may be important orinteresting combinations of variables. Another disadvantage is that the fabricationof discrete libraries can depend upon complex, often expensive, equipment.

An alternate library design and fabrication strategy, and the one we focus on inthis article, is the continuous gradient [2]. In this scheme, diversity is created byfabricating a specimen that gradually and continuously changes in a given param-eter as a function of position (or as we will discuss later, as a function of time).Two or three continuous gradients can be combined in a single system. An illus-tration of a binary continuous gradient library can be seen in Fig. 1. Because theyare continuous, and there are no “gaps” in the parameter space, gradient librariespresent clear advantages for comprehensively examining and mapping the effect ofthe graded properties. For example, a binary gradient library exhibits every possiblecombination of the two graded parameters. As such, the gradient library design isexcellent for mapping property correlation, and for identifying optimal conditionsor critical phenomena that may exist only over a small range or at a specific param-eter combination. Indeed, for materials scientists, gradient libraries can be a naturalexperimental design, since they are quite similar to the phase diagrams they use torepresent two and three parameter systems.

A key aspect of gradient libraries is that they reside on a single substrate, and thishas several advantages. Foremost, gradient libraries yield an entire set of systematicresults from a single compound experiment. In this sense, they are “self-reporting”,


Fig. 1 Illustration of a 2Dcontinuous gradientcombinatorial library thatexhibits gradual andsystematic changes in twovariables

Pro

per

ty 1

Property 2

meaning that they can illuminate trends and express key results without extensiveanalysis (our example of a single specimen phase diagram, discussed below, willillustrate this point). Moreover, because the entire library undergoes identical pro-cessing, “sample to sample” errors (inherent to combining single measurements onindividual specimens) can be reduced. Finally, gradient libraries can often be pro-duced with simple, inexpensive equipment, which makes this approach accessibleto academic laboratories and small companies.

3 Gradient Library Fabrication Methods and ApplicationExamples

Gradient libraries generally assume a planar form, and are supported by a substrate.Often the library is a material deposited onto the substrate as a thin coating, or it canbe achieved through chemical modification of the substrate itself. This geometrynaturally gears gradient libraries for the examination of thin films, coatings and ofthe interfacial properties that govern these systems. At NIST, our goal was to createa suite of gradient library fabrication technologies that would be a combinatorialplatform for examining polymer thin film physics phenomena including wetting andstability, blend phase behavior, self-assembly, and confinement effects.

3.1 Flow Coating: Polymer Film Thickness Gradients

Film thickness can govern the morphology, stability, and surface-chemical expres-sion of polymeric thin films. NIST researchers developed a process for producinggradients, termed flow coating, which is a modified blade-casting technique [3–5].Flow coater instrumentation and the flow coating process are illustrated in Fig. 2.To create the library, a dilute solution of polymer in solvent (1–5% mass fraction)is injected into the gap between a doctor blade positioned over a flat substrate


Fig. 2 Flow coater forcreating polymer gradientthickness libraries: A doctorblade; B polymer solution; Csubstrate; D thickness library;E x-translation stage; Fy-translation stage (forcharacterization).(Reproduced with permissionfrom [13])

(e.g., silicon wafer) mounted on a computer-controlled translation stage. Thestage/substrate is accelerated beneath the stationary blade in the x-direction asshown in Fig. 2. As the stage accelerates, increasing amounts of solution are de-posited along the substrate. Subsequent solvent evaporation results in a gradientpolymer film thickness library. As demonstrated by Stafford et al. [5], the range andslope of the thickness gradient can be precisely tuned through the stage velocityprofile, solution concentration, and gap height.

Films produced via flow coating can be from ≈20 to ≈1mm in thickness. A typ-ical library will double in thickness over about 40 mm in length. In the range of50–600 nm, libraries exhibit constant slopes of ≈1–10nmmm−1, depending uponprocessing parameters. In the NIST instrument, thickness gradients are character-ized via spot interferometry. In this scheme, stacked translation stages (includingthe stage used to produce the specimen) raster the sample beneath the interfer-ometer footprint, resulting in a 2D map of film thickness. With careful instrumentconstruction and operation [5], thickness libraries created via flow coating are lin-ear along the x-direction, and level along the y-direction (to about 3% of the averagefilm thickness at a point x), but 2D thickness characterization may be necessary forthe most quantitative combinatorial analysis.

When used with polymer solutions in the few percent by mass range, the flowcoating instrument creates films that are comparable in thickness and roughnessto those created by spin casting, i.e. ranging from ≈10 to 500 nm thickness, withroot-mean-squared roughness ≤1nm. However, it should be noted that spin cast-ing typically drives solvent from the system much faster than with flow coatingand this can affect the film morphology [6]. Recently, de Gans and coworkers [7]demonstrated a “sector” spin casting technique for creating discrete polymer filmlibraries. In this technique, a metal template is used to divide a round substrateinto pie-slice shaped sectors, into which a series of polymer solutions can be de-posited and cast. By varying spin speed and solution concentration, a discrete filmthickness library could be built; moreover, the method can be used to cast discretecompositional libraries of, for example, polymer blends. Another route for discrete


film library fabrication is ink jetting, which has been explored by several groups [8–11]. As with sector spin coating, this technique can be used to create compositionalspreads. However, ink jetting does not seem to offer the deposition control necessaryto produce well-behaved thickness libraries, since ink jetted spots typically exhibitirregular thickness profiles due to “coffee ring” and other drying effects.

3.2 Gradient Hot Stage: Temperature Processing Libraries

Polymer thin film properties are often governed or modified by high-temperatureannealing and processing, with most phenomena (phase transitions, dewetting, melt-ing etc.) occurring below about 300 ◦C. Accordingly, a temperature gradient withmodest range can create a useful map of the effect of temperature on polymer filmlibraries. This concept has been examined in the literature [12], and in past decadesseveral companies have used this concept to produce gradient hot-stages for appli-cations that include melting-point determination. In recent years, NIST researchersdesigned a gradient hotstage with the aim of producing a flexible instrument thatcould accommodate libraries of various lengths, and that had a tunable temperatureprofile [13–15].

Figure 3 illustrates the NIST gradient hot stage design. The instrument con-sists of an aluminum sample platen (10cm× 15cm× 0.5cm) perforated with twoslots (along x). Two aluminum blocks, fitted with heating/cooling channels, areattached to the bottom of the platen through the slots. This set-up enables con-trol of the inter-block distance, which allows the positions of the heating/coolingsources to be matched with length of the specimen library. The blocks hold cylin-drical heating cartridges or accommodate plumbing for fluid-mediated cooling.

Fig. 3 NCMC gradient hot stage: A sample platen; B slots for mounting/positioning of heat-ing/cooling blocks; C block with channel for heating/cooling element with thermocouple portsfor temperature control; D thermocouple ports for gradient characterization; E ceramic blocks formounting hot stage; F block with cylindrical heating element and thermocouple installed. (Repro-duced with permission from [13])


Proportional–integral–derivative (PID) temperature controllers maintain the blocktemperatures, measured through integrated thermocouples. By heating (or cooling)each block to a constant temperature, a temperature gradient is produced across theplaten. Ceramic supports insulate the device, so it can be mounted on, for example,a microscope translation stage or other observation instrument. The range and slopeof the temperature gradient are tailored through the block distance and the temper-atures of the heating/cooling sources. Remarkably, for modest temperature ranges(≈200 ◦C) the temperature profile is linear [14] along the gradient (x-direction),and level perpendicular to the gradient (y-direction), so measurements of the tem-peratures at the ends stage are sufficient to characterize the gradient. However, inthe NIST instrument, thermocouple ports drilled into the platen edge enable tem-perature measurements along the gradient. Using this device, typical temperaturegradients span intervals of about 100◦C over a total range of room temperature toabout 300 ◦C.

In conjunction with both in situ and ex situ automated measurements, a gradienthot stage can be a powerful tool for examining the role of temperature on film mi-crostructure, morphology development kinetics, phase transitions and performance.For example, Lucas et al. [15] recently used the NIST gradient hot stage to mapthe effect of temperature processing on the structure and performance of organicsemiconductor films. In this study, polythiophene thin films were annealed on atemperature gradient that crossed the bulk liquid-crystal transition of the material.The resulting film library exhibited a range of morphologies across this transition,which could be observed via atomic force microscopy (AFM) conducted along thelibrary. In addition, this team employed an automated probe station to measure thefield effect mobility along the specimen, resulting in a map of the material’s elec-tronic performance. With this strategy, and using a single specimen, the authors wereable both to identify the annealing temperature that gave optimal performance, andto determine how the mobility was correlated with film microstructure. In a similarscheme, Eidelman and coworkers [16] used the gradient hot stage in conjunctionwith automated Fourier transform infrared (FTIR) microspectroscopy and high-throughput adhesion testing to map correlations between curing temperature, degreeof curing and surface tack of model epoxy adhesive formulations. By combining atemperature gradient, flow coating, and automated optical microscopy, Beers et al.[14] were able to examine simultaneously the roles of temperature and thicknesson the crystallization rate of isotactic polystyrene. In this study, flow coating wasused to create a film thickness library of the polymer, which was placed orthogonalto a temperature gradient. Automated micrographs were collected across a grid ofpoints on the library, each of which represented a different combination of thicknessand crystallization temperature. Time sequences of crystallite growth were built byrepeating the cycle of micrograph acquisition over a few hours, which yielded a 2Dmap of crystallization rates as a function of thickness and temperature. In addition,analysis of the library via AFM and optical microscopy allowed the researchers toobserve several thickness dependent transitions in crystallite morphology.


3.3 Surface Energy and Surface Chemistry Libraries

A key factor that governs the behavior of polymer thin films is its interactionwith an underlying substrate. In combination with other factors, surface energyand chemistry of the substrate can cause film instability and dewetting, shifts inthermodynamic and morphological transitions, changes in nano-domain orientation,and modifications in the expression of chemical moieties at the substrate/film in-terface and free film surface. In addition, surface energy/chemistry can affect themobility, growth and morphology of adsorbed cells in biological systems. NISTresearchers have developed two main strategies for creating libraries of substratesurface energy/chemistry, which are useful for screening the effect of this factor onthe behavior of overlying films and other materials. First, we will discuss the useof graded ultraviolet light–ozone (UV–ozone) exposure to fabricate surface energylibraries, and some of the combinatorial studies that resulted from this capability.Then, in the next section, we will consider more sophisticated surface chemistrylibraries fabricated through surface-initiated polymerization, and the use of thesegraded polymer “brush” layers for high-throughput analysis.

It is well known that exposure of organic molecules to ozone generated fromultraviolet–light (UV–ozone) can cause a variety of oxidative reactions – thisis the basis of UV–ozone surface cleaning devices. For alkyl chain molecules,UV–ozonolysis leads first to the formation of oxygen containing moieties, typi-cally starting at the chain ends, with the eventual consumptive oxidation of themolecules at long exposures. Since the degree of oxidation is dependent uponthe UV–ozone dosage, this process can be harnessed to create a gradient surfaceenergy library. There are several ways to create the gradient in UV–ozone exposureneeded to accomplish such libraries. In one strategy, NIST researchers [17, 18]used a UV–ozone flood source (185 and 254 nm light) to illuminate a planar sub-strate through a graded neutral density filter, which systematically decreased theamount of transmitted light as a function of position. The substrate was a nativeoxide-terminated silicon wafer, treated with a self-assembled monolayer (SAM) ofn-octyldimethylchlorosilane (ODS).

As shown in Fig. 4a for an ODS treated substrate, the library exhibits a system-atic change in water contact angle along its length. Roberson and coworkers [17]also used more sophisticated contact angle measurements to yield the polar anddispersive parts of the total surface energy (Fig. 4b). These measurements demon-strate that the UV–ozone treatment changes the polar part of the surface energy,while leaving the dispersive part relatively unchanged. Time-of-flight secondary ionmass spectrometry measurements across the library show that the UV–ozonolysisgradually imparts the hydrophobic, methyl-terminated, ODS SAM layer with a va-riety of oxygen containing end-groups, but these were primarily –COOH terminatedspecies. The growing number of –COOH terminated species along the library resultsin its increasing hydrophilicity (i.e. lower water contact angles) as shown in Fig. 4.

More recently, NIST researchers [13, 19] developed a device to more preciselygenerate surface energy libraries using OV-ozonolysis. Pictured in Fig. 5, this de-vice achieves graded UV–ozonolysis through a computer-driven translation stage,


Fig. 4 Water contact angledata (a) and surface energydata (b) from a surfaceenergy library producedthrough the gradedUV–ozonolysis of an ODSself-assembled monolayer onsilicon. (Reproduced withpermission from [17])

80

60

Con

tact

Ang

le, q

°

40

20

0

80 σsp

σsd

σs

b

a

60

Surf

ace

Ene

rgy,

s (m

N/m

)

40

20

0

0 10 20 30 40 50x, mm

0 10 20 30 40 50

x, mm

θH2OθCH2I2

lightsourcehousing

light source

sample stage

source height adjustment

actuator

Fig. 5 Illustration of NIST Gradient UV–Ozone device for generating surface energy librarieson substrates functionalized with hydrophobic SAM species. The sample stage accelerates thespecimen (blue) beneath a slit-source of UV light. (Reproduced with permission from [12])


which accelerates the silanized substrate beneath a 185 nm/254 nm UV wand-sourceprojected through a 2 mm-wide slit aperture cut into the cylindrical lamp housing.The gap between the aperture and the substrate is controlled through a micro-positioner incorporated into the fixed lamp mount. Using these exposure-mediatedmethods on ODS treated silicon, surface energy gradients can span any interval be-tween 20mJm−2 and 75mJm−2, over a tunable length of 1–5 cm. Water contactangles typically span 100◦ to less than 10◦.

The main advantage of this approach is the ability to determine the stage acceler-ation profile via computer control [19]. As opposed to a graded density filter (whichrepresents a single static exposure “function”) any mathematical function can be fedinto the stage motion routine. This enables users to tune the length, steepness andshape of the surface energy profile in the library. Indeed, this sort of control canbe used to create a well-behaved surface energy gradient profile, which can ease itsapplication in combinatorial screening. For example, this capability can be used tocreate libraries that have a linear surface energy gradient, rather than the sigmoidalsurface energy profile that results from a linear exposure function (as in Fig. 4). Inthis sense, this method also presents an advantage over techniques for creating sur-face chemistry gradients via diffusion mediated deposition of SAM species (see forexample [20–22]), since these techniques also result in a steep, sigmoidal gradientprofile. However, diffusion techniques and controlled immersion methods [23] offerthe possibility of creating more chemically diverse mixed SAM gradient librariesvia the simultaneous graded deposition of two SAM species [24, 25], or through abackfill sequence [26, 27].

The graded UV–ozonolysis approach to surface energy library fabrication is apowerful tool for examining the role of substrate-polymer interactions on film phe-nomena such as dewetting and block copolymer self-assembly. Ashley et al. [28],used this approach to determine the surface energy dependence of the stability anddewetted morphology of polystyrene films. In conjunction with an orthogonal tem-perature gradient, and automated optical microscopy, this team was rapidly able tomap the dewetting behavior of five polystyrene specimens that varied in their molec-ular mass. The library approach enabled the team to observe the surface energy andtemperature bounding conditions for film stability, since the dewetted portions ofthe film could be easily screened. Indeed, the authors discovered that, for the rangeof molecular mass they considered, these boundary conditions could be collapsedto a common “master curve”, which suggests that a universal surface energy andtemperature-dependent dewetting behavior is exhibited by this system.

A number of researchers have used surface energy libraries to examine theself-assembly of block copolymer species in thin films. It is well known thatsubstrate-block interactions can govern the orientation, wetting symmetry and eventhe pattern motif of self-assembled domains in block copolymer films [29]. A sim-ple illustration of these effects in diblock copolymer films is shown schematicallyin Fig. 6. However, for most block copolymer systems the exact surface energy con-ditions needed to control these effects are unknown, and for many applications ofself-assembly (e.g., nanolithography) such control is essential.


a b c d

A

L0/2

B

L0L0

h = mL0

h = (1 1/2)L0 Surface perpendicular domains

hydrophobicsymmetric asymmetric hydrophilic

Fig. 6 Illustration of surface energy effects on the self-assembly of thin films of volume symmetricdiblock copolymer (a). Sections b and c show surface-parallel block domains orientation that occurwhen one block preferentially wets the substrate. Symmetric wetting (b) occurs when the substrateand free surface favor interactions with one block B, which is more hydrophobic. Asymmetricwetting (c) occurs when blocks A and B are favored by the substrate and free surface, respectively.For some systems, a “neutral” substrate surface energy, which favors neither block, results in a self-assembled domains oriented perpendicular to the film plane (d). L0 is the equilibrium length-scaleof pattern formation in the diblock system

≈60 nm

≈80 nm

thic

knes

s g

rad

ien

t

substrate surface energy gradient≈52 mJ/m2 ≈32 mJ/m2

3 mm

h ~ 3.5 L0

h ~ 3 L0

h ~ 4 L0

Asymmetric Wetting

smoothSurface-Perpendicular

Domains

Sym

met

ric

wet

ting

Fig. 7 2D thickness-surface energy gradient library for mapping the effects of these parameterson the self-assembly of PS-b-PMMA block copolymer thin films. See text for a full description.L0 is the equilibrium self-assembly period and h is the film thickness. Dashed white lines delineatethe “neutral” surface energy region, which exhibits nanostructures oriented perpendicular to thesubstrate plane. (Derived from [18] with permission)

To address this problem, Smith and coworkers [18] combined flow coatingand surface energy gradients to generate a single specimen that maps the effectsof film thickness and surface interactions on the self-assembly of polystyrene-b-poly(methyl methacrylate) (PS-b-PMMA) thin films. This thin film library isshown in Fig. 7. As demonstrated by this photograph of the specimen, the for-mation of microscopic “island and hole” structures, which make areas of the filmhazy, make the library “self-reporting” via simple visual inspection. These surfacefeatures only occur when the block domains are oriented parallel to the surface,and they have a specific thickness-dependence, i.e. they disappear when the filmthickness is an exact integral (or n + 1/2 integral) of the equilibrium period (L0)of the diblock self-assembly. Examination of the thickness-dependence enabled theauthors to determine the wetting symmetry of the surface-parallel self-assemblyin these areas (see Fig. 6 caption). Moreover, the range of “neutral” substrate


surface energy is clearly indicated by the vertical band of film in which there is alack of hazy bands regardless of film thickness. This range of surface energy causesthe domains to orient perpendicular to the film surface. In this study, Smith et al.used the method to examine rapidly the molecular mass dependence of the width ofthe neutral surface region, which is a measurement of the diblock’s susceptibilityto surface interactions. Moreover, a separate gradient analysis of the thickness-dependence of the “island and hole” structures in diblock films yielded observationsof a new labyrinthine form of this surface phenomenon [30].

Other teams used the surface energy gradient approach to examine the morecomplex thin film self-assembly of three component triblock copolymers. For exam-ple, Ludwigs et al. [31] used the technique to screen for surface energy dependentmorphological shifts in a PS-b-poly(vinylpyridine)-b-poly(tert-butyl methacrylate)triblock films. Remarkably, in a single library experiment, the authors observedthat the system exhibited a perforated lamella motif in thin films regardless of thesubstrate surface energy. This finding has interesting implications for lithographicapplications, since it indicates that this laterally structured nanopattern can beformed on almost any smooth substrate material. More recently, Epps and cowork-ers [32] performed a similar study of a polyisoprene-b-PS-b-poly(ethylene oxide)triblock films. As with the work of Ludwigs et al., this examination showed that,regardless of the surface energy of the substrate, the triblock formed a different mor-phology in thin films to that in the bulk, in this case three-layered surface-parallellamella. However, the authors observed that, at long annealing times, the lamellarfilms unexpectedly dewet the substrate over a specific range of surface energy. Us-ing a high-throughput technique developed to pluck film specimens from the library[33] the team could examine the buried film–substrate interface via automated X-ray surface spectroscopy methods. Using this data, the unexpected phenomenon wasdetermined to be a form of surface energy dependent autophobic dewetting that hadnot been observed previously. In both of these studies, it is unlikely that these keyobservations could have been made as readily without the comprehensive scope ofthe gradient library approach.

The UV–ozone gradient exposure approach can also be used to fabricate morecomplex libraries. For example, Julthongpiput and coworkers [34, 35] employedthe technique to create libraries that exhibit a graded chemical micropattern. Asillustrated in Fig. 8, these combinatorial test substrates consist of a pattern ofmicron-scale lines that exhibit a continuous gradient in surface energy differencesagainst a constant surface energy matrix. On one end of the specimen the lines arestrongly hydrophobic while the substrate matrix is hydrophilic SiO2. The lines be-come increasingly more hydrophobic towards the other end of the library until theyare chemically indistinguishable from the matrix. The library is fabricated througha vapor-mediated soft lithography [34] of an ODS SAM which is then treated to agraded UV–ozonolysis with the device shown in Fig. 5. The library design includestwo calibration strips that express the changing and static surface energy of theSAM pattern lines and matrix respectively. Accordingly, the surface energy differ-ences along the patterned region can be determined by contact angle measurementsalong the calibration strips.


static γ Hydrophilic MatrixCalibration field

Gradientcalibration field

Gradientmicropattern

HydrophilicHydrophobicvariable γ

Fig. 8 Illustration of a gradient micropattern library. The central band of the library exhibits amicropattern that gradually changes the chemical differences between the striped domains and thematrix until the surface is chemically homogeneous. The bands on the top and bottom of the libraryare the calibration fields for static matrix and gradient respectively. γ is surface energy

Decreasing Pattern Surface Energy Differences

Fig. 9 Optical micrographs of dewetted polystyrene droplets collected from points along a chem-ical gradient library. Scale bar (red) is approximately 20μm. As discussed in the text, this librarywas used to examine the transition between pattern directed dewetting (left micrographs) andisotropic dewetting from a homogeneous surface (right micrograph). (Reproduced with permis-sion from [35])

The gradient micropattern library is a unique combinatorial tool for examiningthe effects of substrate chemical heterogeneity on surface, interface and thin filmphenomenon. Julthongpiput et al. [35] recently demonstrated this library as a meansto determine how the chemical differences between the stripes can drive the rup-ture, dewetting and patterning of overlying polystyrene thin films. In this study,automated optical microscopy was employed to collect 1,700 contiguous opticalmicrographs of polystyrene droplets that formed as a result of dewetting from thegradient micropattern substrate. A small, representative selection of these images isshown in Fig. 9. Automated image analysis was used to examine the droplet arrange-ments, in particular their registry with the underlying pattern. Through this data, theteam could determine the range of surface energy differences that resulted in “pat-tern directed” dewetting, i.e. film instability, rupture and droplet alignment causedby the underlying stripes. In particular, the library showed that chemical differencesbetween 14mJm−2 and 20mJm−2 resulted in the best droplet alignment. Moreover,the library data showed that when the pattern chemical differences dropped below7mJm−2, the droplets had an isotropic arrangement indicative of dewetting from ahomogeneous surface. This key observation, which precisely illuminates the mini-mum surface-chemical heterogeneity needed to induce film dewetting, would havebeen very difficult to make using traditional “single specimen” methods.


Another variation and application of the UV–ozone surface energy library wasdemonstrated recently by Gallant et al. [36]. In this work, the surface energy li-brary is used as a facile platform for further modification by the “click” chemistry[37] route. In particular, the graded –COOH terminated SAM molecules producedby the UV–ozonolysis are reacted with a bifunctional cross-linker terminated withboth amino and alkyne functional groups. The result is a library that has an increas-ing density of SAM chains with alkyne functionality. Through the click chemistryscheme, the alkyne can react with a huge variety of azido-derivatized biofunctionalmolecules. Accordingly, this versatile scheme enables continuous gradients in thegrafting density of a number of bioactive species. The authors demonstrated thisapproach by creating a library that continuously varied the concentration of surface-bound RGD peptide molecules. This gradient was used to measure the effect ofRGD density on cell adhesion and morphology. Using automated fluorescence mi-croscopy, the team was able to measure cell behavior under a huge number of RGDconcentrations, and determine the RGD densities that resulted in the most numberof adhered cells, and the most extensive cell spreading. Such data is critical whendesigning surfaces for cell scaffolds and other biomaterials applications.

The general approach of graded radiation exposure can also be used to examinelight driven processes such as photopolymerization [19]. For example, Lin-Gibsonand coworkers used this library technique to examine structure-property rela-tionships in photopolymerized dimethacrylate networks [38] and to screen themechanical and biocompatibility performance of photopolymerized dental resins[39]. In another set of recent studies, Johnson and coworkers combined gradedlight exposure with temperature and composition gradients to map and model thephotopolymerization kinetics of acrylates, thiolenes and a series of co-monomersystems [40–42].

3.4 Gradient Polymer Brush Libraries

While UV–ozone-generated surface energy libraries are simple to implement, theypose limitations in terms of both stability and chemical diversity. ChlorosilaneSAMs modified via UV–ozonolysis are susceptible to degradation under light ex-posure, oxygen, humidity, and high-temperature, and thus must be used within afew hours of fabrication. Moreover, without further modification (as in the examplefrom Gallant and coworkers above), this route generally results in a gradient only be-tween –CH3 and –COOH functionalities. In order to create more robust and diversesurface chemistry gradients, NIST researchers turned to surface-initiated polymer-ization (SIP) techniques [43, 44] to create libraries of grafted polymer “brushes”that would systematically change in their molecular composition and architecture.In this endeavor, the pioneering work of Genzer and coworkers [45–48] provided ex-amples to build upon, including the creation of gradient libraries of polymer brushlength and grafting density.

SIP involves the growth of a polymer chain from an initiator moiety that hasbeen covalently tethered to a surface. The advent of this reaction approach, along


with advances in polymer synthesis routes, have enabled the creation of denselypacked polymer “brush” layers that can exhibit a huge variety of macromolecularcompositions, architectures and functional groups [49, 50]. Because this covalentlybound layer can result in overlapping polymer coils, surface coverage is enhanced.Moreover, some polymer types and architectures offer the possibility of generat-ing surfaces with advanced functionality including, chemical switching, reversiblewetting and bioactivity [51, 52]. The potential chemical diversity and technologicalpromise of polymer brushes present opportunities for new library fabrication meth-ods, and for applications of combinatorial techniques. In the following passages, wewill discuss how NIST researchers examined both of these opportunities.

SIP-driven polymer brush library fabrication leverages the fact that the poly-merization initiation species are permanently bound to the substrate. Since theinitiators are tethered, controlled delivery of monomer solution to different areasof the substrate results in a grafted polymer library. In NIST work, initiators boundvia chlorosilane SAMs to silicon substrates were suitable for conducting controlledatom transfer radical polymerization (ATRP) [53] and traditional UV free radicalpolymerization [54, 55]. Suitable monomers are delivered in solution to the surfacevia microfluidic channels, which enables control over both the monomer solutioncomposition and the time in which the solution is in contact with the initiatinggroups. After the polymerization is complete, the microchannel is removed fromthe substrate (or vice versa). This fabrication scheme, termed microchannel confinedSIP (μ-SIP), is shown in Fig. 10. In these illustrations, and in the examples discussedbelow, the microchannels above the substrate are approximately 1 cm wide, 5 cmlong, and 300–500 μm high.

a

b

Monomer solution in syringe pumpRemovable Microchannel

Block Copolymer Library

Monomer solution flow

Monomer A

Monomer B

Poly (A-stat-B) Gradient

Hompolymer MolecularWeight Library

Initiator-functionalizedsurface

Microfluidic static mixer

Fig. 10 Illustrations of the microchannel confined surface-initiated polymerization (μ-SIP) routefor producing gradient polymer brush libraries: a route for making polymer molecular weight andblock copolymer libraries; b route for making statistical copolymer libraries. Red arrows show theflow of monomer solution from a syringe pump used to gradually fill the microchannel. See textfor details


As illustrated in Fig. 10a, and described by Xu and coworkers [56], the mostbasic implementation of μ-SIP involves a gradual filling of the microchannel withmonomer solution under polymerization conditions. Depending on the rate of thepolymerization and the desired library design, the microchannel is filled using acomputer-controlled syringe over 5–40 min. Accordingly, beginning from the mi-crochannel feed, the substrate is exposed to the monomer solution for a decreasinglength of time, which is equivalent to a decreasing polymerization period along thelibrary. This process results in a grafted polymer library that gradually decreasesin its molecular mass and is evidenced by decreasing film thickness, as demon-strated in Fig. 11a for a poly(N,N-dimethylaminoethyl methacrylate) (PDMAEMA)homopolymer [56]. If a living polymerization route (such as ATRP) is used to growconstant-length polymers from the substrate, a similar process can be employed witha second monomer solution to fabricate a block copolymer library in which the sec-ond block gradually decreases in its length. As shown in Fig. 11b, Xu and coworkers[57, 58] demonstrated this technique by fabricating a series of grafted block copoly-mer libraries consisting of poly(n-butyl methacrylate) (PnBMA) and PDMAEMA.An application of these libraries will be discussed below. A variation of this ap-proach can be used to create a gradient in polymer grafting density. As demonstratedby Mei et al. [59], this involves creating a gradient in surface-bound initiator con-centration, which is achieved by gradually introducing initiator-SAM solution alongthe substrate. Subsequent immersion in the monomer solution results in a librarythat systematically varies the lateral spacing of tethered polymer chains. A simi-lar technique for creating grafting density libraries has been published by Wu andcoworkers [47].

As illustrated in Fig. 10b, a more sophisticated statistical copolymer gradient li-brary can also be fabricated through the μ-SIP method [60]. The key to this libraryis a microfluidic mixer, positioned between two monomer solution feeds and themicrochannel. The mixer serves to combine a ramped flow of these solutions inwhich the relative amount of one solution is decreased and the other increased overtime. The result of this ramped input is that the microchannel is filled with a gradi-ent monomer solution composition. Because of the narrow height dimensions of thechannel, cross diffusion of the monomers is suppressed, and the solution gradientcan persist over a few hours. Accordingly, during a period of polymerization, themonomer concentration profile is transferred to grafted polymer chains on the sub-strate. The resulting statistical copolymer brush gradient gradually changes fromnearly 100% of one monomer to 100% of the other along the library. Composi-tion data from a PnBMA-s-PDMAEMA statistical copolymer library, collected vianear edge X-ray absorption fine structure (NEXAFS) spectroscopy, can be seen inFig. 11c [60]. This achievement is exciting since it represents a way of producingsurface chemistry libraries that exhibit both the reliability and enhanced coverageof covalently bound polymer brushes, and the potential for creating chemical gra-dients that exhibit the extensive chemical and architectural functionality availablefrom advanced polymerization routes. In addition, a key feature of grafted statisti-cal copolymers is that they present stable and intimate blends of disparate chemicalspecies. Since the different monomers are bound within the same chains, the system


Distance from Microchannel Feed

Filled: Brush thickness after both blocksOpen: Brush thickness after first block

35

30

25

20

Thi

ckne

ss /

nm

15

10

5

00

c

b

a

20 40 60

Distance / mm

0.06

0.6

0.4

0.2

0.0

0 10 20 30

σNC

40

0.5 mm10.5 mm15.5 mm20.5 mm25.5 mm30.5 mm35.5 mm40.5 mm45.5 mm

Position (mm)

Nitr

ogen

Inte

nsity0.05

0.04

0.03

Par

tial E

lect

ron

Yie

ld

0.02

0.01

0.00380 390 400 410 420

Photon Energy (eV)

*

Fig. 11 Data from polymer brush libraries generated by μ-SIP: a photograph of molecular weightgradient of grafted PDMAEMA. The entire library is approximately 50 mm long. The brush thick-ness ranges from approximately 60 to 0 nm from left to right. (Reproduced with permission from[56]); b thickness data (ellipsometry) from a PnBMA-b-PDMAEMA block copolymer library afterthe growth of the PnBMA block (open symbols) and the PDMAEMA block (closed symbols). (Re-produced with permission from [58]); c NEXAFS data along PnBMA-s-PDMAEMA statisticalcopolymer gradient library. The decreasing nitrogen edge electron intensity signal demonstratesthat the DMAEMA segment content of the statistical copolymer systematically decreases alongthe library. (Reproduced with permission from [60])


will not undergo lateral clustering typical of mixed homopolymer brushes [61], orsegment segregation seen in block copolymer brushes [45, 58], both of which cancause spatial chemical heterogeneities on the scale of 10 nm or more. Accordingly,if local surface-chemical homogeneity is needed from a surface chemistry gradientlibrary, statistical copolymer gradients present an advantage.

Grafted homopolymer and copolymer gradient libraries have a tremendouspotential for the high-throughput examination of both brush properties and ap-plications involving grafted polymer layers. For example, NIST researchers Xuet al. [57, 58] used such libraries to examine the ability of grafted block copoly-mers to change their surface segment expression under different environments.This “switching” behavior can be harnessed to create “smart” coatings and surfacesthat change their wetting, adhesion, and other properties in response to environ-mental triggers. The NIST study examined environmental response of the seriesof PnBMA-b-PDMAEMA block copolymer libraries shown in Fig. 11b. In thissystem, the PDMAEMA (top) block segments are preferentially solvated by water,while hexane is a preferential solvent for the PnBMA (bottom) blocks. Basically,the gradient experiment involved an assessment of the expression of PnBMA andPDMAEMA segments at the surface of the brush after the library was treated withwater and then hexane. Water contact angle measurements along the library wereused to estimate the degree of segment surface expression, based on the known equi-librium water contact angles for the pure polymer species (about 90◦ for PnBMAand approximately 65◦ for PDMAEMA). The results of these experiments for threelibraries, each with a different PnBMA block length, can be seen in Fig. 12. In thisplot, the brush is shown to “switch” its surface expression of segments betweenthe block species when the measured contact angle changes due to water or hexaneexposure. If the contact angle remains the same, it indicates that the segments wereunable to significantly rearrange. The power of the gradient approach to this systemis that it clearly outlines how molecular parameters govern the diblock switchingbehavior. In addition, it provides a view of narrow windows of optimal responsethat would be quite difficult to observe in single specimens. The library showsthat longer PDMAEMA blocks suppress switching, while longer PnBMA blocksenhance the system’s ability to rearrange. In addition, the library illuminates thenarrow ranges of molecular architecture that result in the maximum changes in seg-ment expression at the surface. For example, for the data shown in red this optimalswitching occurs in a window of top block thickness that is only 4 nm wide.

Grafted polymer libraries were also used by NIST researchers to achieve a va-riety of other measurements. For example, Mei and coworkers [59] leveraged theirgradient in polymer grafting density to assess brush biocompatibility. In particu-lar, they used this approach map the effect of poly(2-hydroxyethyl methacrylate)grafting density on the level of fibronectin adsorption and subsequent cell bind-ing. The library enabled the team to determine rapidly the complex correlationsbetween polymer grafting density, fibronectin coverage and cell adhesion, as wellas the optimal surface conditions for cell proliferation. In another emerging ex-ample, Patton and coworkers [62] demonstrated how μ-SIP can be leveraged tomeasure rapidly and reliably copolymer reactivity ratios, which link the composition


Water

4.5 nm9.8 nm

14.1 nm

Bottom block PnBMA thickness:

PDMAEMA (top) block thickness (nm)

Wat

er C

onta

ct A

ngle

(°)

Noswitch

Filled symbols: After hexaneOpen symbols: After water

switchNo

Switch

Hexane

95

a

b

90

85

80

75

70

65

600 2 4 6 8 10 12

Fig. 12 Surface expression of block segments in block copolymer gradient libraries after treat-ment to two solvents. See text for details: a illustration of surface expression of PnBMA (black)and PDMAEMA (blue) block copolymer brush segments after water and hexane treatments; bwater contact angle data from three PnBMA-b-PDMAEMA block copolymer gradient librariesafter hexane (filled symbols) and water (open symbols) treatments. (Derived from [58] with per-mission)

of mixed monomer polymerization solutions to the composition of copolymerspecies. Knowledge of reactivity ratios is extremely important for synthesizingpolymers with tailored composition and architecture, but they are difficult and time-consuming to measure. The researchers showed that by measuring the compositionof the monomer solution in the microchannel (e.g., via a fiber optic Raman spec-troscopy probe) and correlating it to the composition of the statistical copolymerbrush created on the surface (measured by X-ray photoelectron spectroscopy, XPS),reactivity ratios could be determined. While the published study involved a seriesof discrete specimens, the extension of the measurement approach to gradient li-braries is straightforward. This team is currently establishing protocols to reliablyachieve this by combining Raman spectroscopy along a μ-SIP solution gradient andautomated XPS data collected from a gradient statistical copolymer library of thetype illustrated in Fig. 10b.


3.5 Polymer Blend Composition Gradients

In “hard” materials, such as metals or oxides, the creation of composition gradientsis enabled by the excellent atomic-level mixing inherent to co-sputtering, co-evaporation and other co-deposition techniques [63]. The creation of compositionspreads in polymers can be more difficult, since the size of macromolecules and thehigher viscosity of their solutions can inhibit the proper mixing required for reliablelibraries. At NIST, Meredith and coworkers addressed this challenge by combiningthe flow coating apparatus with an automated syringe-based deposition of blendedpolymer solutions [3]. A schematic of this approach can be seen in Fig. 13. To start,solutions of the polymers (A and B in the figure) to be blended are prepared us-ing a common solvent. These solutions are fed, and then continuously mixed, in acommon vessel (Fig. 13a). The solution feeds are ramped over time such that themixed solution begin with 100% B solution and ends with 100% A solution. As thecomposition of the mixed solution changes, a narrow bore syringe is used to col-lect gradually an aliquot from the vessel, capturing a column of solution that hasa gradient in its composition from top to bottom. The narrow bore of the syringeinhibits mixing long enough so that it can be deposited as a strip onto a flat sub-strate (Fig. 13b). Then the flow coater blade (moving at constant velocity) is usedto spread the strip into a thin film with level thickness (Fig. 13c). When the solventevaporates, the resulting rectangular film library has a gradient in polymer compo-sition that ranges from nearly 100% A to nearly 100% B.

There are some limitations to this technique. First, proper mixing can only beachieved with dilute, low viscosity polymer solutions (perhaps a few percent poly-mer by mass), so the final films are at most a few hundred nanometers thick. Thiscan be a problem if confinement will create undesired effects in the blend behavior.

B-righ

BA

A-righ

substrate

film

motion st

age

knife

S

WI

compositiongradient column

deposit stripe spread film

∇fB

∇fB

∇fB

a b c

Fig. 13 Illustration of a method for producing polymer blend composition gradient libraries. A andB are the polymer solutions to be blended. ϕB is the relative volume concentration of the B polymersolution. See text for details. (Reproduced with permission from [3])


Temperature gradient

PolystyreneC

ompo

sitio

n G

radi

ent

PVME

Hota bCold155

145

T(°C)

125

115

105

øps0.0 0.2 0.4 0.6 0.8

Fig. 14 Creation of a single specimen polymer blend phase diagram from orthogonal polymercomposition and temperature gradients. The polymers are polystyrene and poly(vinyl methyl ether)(PVME): a composition library placed orthogonal to a temperature gradient; b completed gradientlibrary polymer blend phase diagram. White points are data derived from traditional measurementfor comparison. See text for details. (b reproduced with permission from [3])

In addition, the method only works for polymer pairs that can be blended in a com-mon solvent. Nevertheless, this can be an extremely powerful method for assessingthe behavior and performance of polymer blends. For example, Meredith et al.demonstrated the application of this approach to the high-throughput analysis ofpolymer blend phase behavior. Using polystyrene (PS) and poly(vinylmethylether)(PVME) as a test system, these authors created a single specimen polymer phase di-agram by placing a PS-PVME gradient library on a gradient hot stage, as shown inFig. 14. After annealing over the range of temperatures across the gradient hot stage,the library developed a hazy region, indicative of phase separated polymer domains.Where the polymer remained mixed, the film retained its smooth, as-cast appear-ance. As shown in Fig. 14b, these hazy and smooth regions delineate the miscibilitygap of the PS-PVME blend system. The resulting polymer blend phase diagramlibrary captures the entirety of this system’s phase behavior in a single specimen.

Several research teams have employed this method for the high-throughput inves-tigation of more complex polymer blend behavior and performance. For example,Karim and coworkers [64] used the method to screen the shifts in polymer miscibil-ity induced by the addition of clay nanoparticles. In another study, a blend systemslated for biomaterials applications was examined by Meredith et al. [65]. In thiswork, after the gradient phase diagram was created and cooled, cells were seededacross the library. A subsequent stain for cell viability indicated at which locationsacross the library cells adhered and propagated. Thus, in a single library experiment,the researchers could create a comprehensive map of which blend compositionsand morphologies were biocompatible. More recently, Simon and coworkers [66]adapted this method to create microporous 3D gradient polymer blend specimensto investigate tissue scaffold materials. Instead of spreading a graded polymer con-centration solution on a flat substrate, the researchers deposited it along a troughfilled with salt crystals. Freeze-drying to remove the solvent and dissolution of the


salt resulted in a porous monolith (roughly 1cm× 1cm× 5cm in dimension) thatgradually changed in its composition from one end to the other. A first applicationof this technique resulted in the identification of the optimal level of iodinated poly-mer additive needed to create contrast in X-ray imaging measurements (e.g., X-raytomography) of polymer tissue scaffolds [67]. In this study, the gradient polymerblend composition library was fabricated so that it gradually increased in the levelof iodinated species along its length. A single X-ray radiography image of the entirelibrary thus provided a comprehensive map of X-ray adsorption levels as a functionof position. Using only two such libraries, they were able to determine the minimallevels of contrast agents required for four X-ray imaging processes. By traditionalmethods, similar optimization would have required preparation of nearly 100 spec-imens, followed by many hours of imaging measurements.

4 High-Throughput Materials Testing: Surfaces, Interfaces,and Thin Films

In order to tailor the function and properties of next-generation coatings and adhe-sives, industry researchers need to understand and control the complex interactionsof material interfaces. However, the properties of interfaces are often difficult tomeasure, since they are complex in their structure and chemistry, and depend on theinterplay between multiple variables. Consequently, high-throughput measurementsof surfaces, interfaces, and thin films are essential for developing structure-propertyrelationships of coatings and adhesives generated using combinatorial strategiessuch as those presented in the previous sections. The NIST program has focused onenabling measurements of intrinsic properties of polymer films such as the Young’smodulus, an extensive property of a material, as well as extrinsic properties suchas adhesion, which depends on a multitude of factors such as modulus, surface en-ergy, and surface roughness. In designing these types of measurement platforms,we discovered that some approaches are amenable to performing highly-parallelmeasurements on combinatorial libraries; thus, we could provide rapid, multiple-point measurements of a particular response. However, other approaches, due thevery nature of the measurement method itself, did not lend themselves to parallelmeasurements; thus, we incorporated high-throughput, single point measurementsof combinatorial libraries into our experimental design. In the following sections,examples of each type of measurement workflow will be highlighted.

4.1 Thin Film Mechanical Properties

Nanotechnology promises to revolutionize a growing set of materials applicationsranging from technology sectors such as semiconductor manufacturing, advancedsensors and coatings, to biomedical sectors such as drug delivery and implant


devices. However, the quest to engineer materials on the nanoscale (e.g., in theform of ultrathin films) is met with the daunting task of measuring the physicaland mechanical properties of these systems. Given that the material properties ofthin films can be drastically different from that of the bulk material, understandingthe mechanical properties of nanofilms is especially critical not only for engineeringrobust fabrication techniques but also for defining application thresholds and oper-ating windows. Maintaining or even improving device performance and reliabilitywhile concurrently shortening overall time-to-market is strongly dependent on theability to rapidly and quantitatively measure the mechanical properties of thin filmsand coatings. At NIST, we developed several combinatorial and high-throughputmeasurement platforms that probe the mechanical properties of thin film libraries.In particular, we incorporated combinatorial libraries into an established methodol-ogy based on deformation of a thin film on a copper grid to investigate crazing andfracture in thin coatings. We also pioneered a new methodology based on surfacewrinkling to rapidly measure the elastic modulus of thin films and coatings. Thesetwo measurement platforms underscore many of the challenges and opportunitiespresented by combinatorial and high-throughput experimental design.

4.1.1 Crazing in Thin Polymer Films

Upon application of strain, polymeric materials can undergo local deformation andyielding processes such as crazing, which leads to the formation of small fibrils andmicrovoids. These fibrils and microvoids effectively increase the fracture toughnessof the material by absorbing energy prior to large-scale cracking in the material[68]. Since many applications of polymers employ thin coatings that are exposed torelatively large stress fields, it is imperative to understand crazing in thin film ge-ometries. The copper grid technique [69] applies a uniaxial strain to a thin polymerfilm mounted onto a ductile copper grid (see Fig. 15). Due to plastic deformationof the copper, a portion of the applied strain is transferred to and remains in the

Fig. 15 a Digital image of a copper grid-supported thin polystyrene film clamped in a uniaxialtensile machine. b Schematic of a thickness gradient film mounted for conducting a copper gridstrain test. c AFM images (height and phase) of a craze tip after deformation. (Derived from [70]with permission)


attached polymer film, even after the sample has been removed from the tensiletesting instrument. This feature allows for the quantitative assessment of crazemicrostructure, craze distribution and film integrity. Observation of specimen de-formation is achieved by monitoring the material suspended across the grid holes.The fracture processes within each grid space act independently and represent indi-vidual experiments. Accordingly, copper grid testing of gradient specimens enablesparallel screening of craze behavior over the parameter spaced embodied by a com-binatorial library. The ability to analyze rapidly multiple combinations of variablesaffecting crazing on a single sample eliminates potential variability and measure-ment error associated with sample preparation, processing and storage, while at thesame time increasing measurement efficiency. Shallow thickness gradients allowcomparatively uniform films to be presented across each grid square, thus yield-ing the equivalent of up to 30 or more different films that can be analyzed underidentical conditions. The grid holes orthogonal to the thickness gradient can pro-vide statistics of the crazing process, or a second gradient, such as film compositionor crystallinity, can be incorporated into the film, thereby greatly increasing theparameter space studied. Using this technique, NIST researchers demonstrated thatthis method provides quantitative characterization of craze dimensions in glassypolymer films. Interestingly, those results indicated that craze widening and micro-necking mechanisms are quantitatively continuous in films with thickness greaterthan 50 nm [70].

4.1.2 Thin Film Modulus Measurements

While the copper grid test captures the crazing and fracture behavior of thin polymerfilms, it does not provide any measure of the fundamental mechanical properties ofthese materials (e.g., the elastic modulus). The most common method for probingthe modulus of thin coatings and films is instrument indentation (nanoindentationor AFM), which has proven extremely valuable in the field of hard materials suchas metallic and ceramic materials. Despite the success of instrumented indentation,there continues to be a number of technical issues impeding accurate indentationmeasurements on thin polymer films, the most notable being the so-called substrateeffect which necessitates that the indentation depth be less than 10% of the totalfilm thickness. Such shallow indentation depths become increasingly impractical ordifficult as the film thickness approaches 100 nm or less. Furthermore, when study-ing polymer films, it is difficult to detect when the indenter establishes contact withthe surface due to the extremely low loads encountered with softer materials (MPato GPa).

To address this measurement need, NIST researchers developed a novel method-ology based on surface wrinkling to assess the mechanical properties of thin poly-mer films [71, 72]. Surface wrinkling occurs upon compression of a bilayer laminatecomprised of a stiff, thin coating supported by a thick, soft substrate. In order tominimize the applied strain energy, the system undergoes a mechanical instabilityhaving a defined wavelength (λ ), which can be related to the elastic modulus of thestiff coating by


Ef = 3Es

(λ

2πhf

)3

(1)

where h is the thickness, E = E/(1− ν2) is the plane-strain modulus (E is theelastic modulus, ν is the Poisson’s ratio, and the subscripts “f” and “s” denote thefilm and substrate, respectively). In nearly all studies to date, NIST researchers andothers have employed crosslinked poly(dimethyl siloxane) (PDMS, E ≈ 2MPa) asthe substrate. Because the substrate modulus (Es) and film thickness (hf) can both beindependently measured by traditional techniques, the wavelength of the wrinklingprovides a window for measuring of the modulus of the stiff, thin coating (Ef).The wavelength of the wrinkling instability can be measured rapidly by a numberof techniques such as laser light diffraction, optical microscopy, or AFM. In thecase of light diffraction, the sample can be rastered across the beam to map out themechanical properties of the entire film, providing rapid analysis of, for example, agradient library. Conversely, if the sample is uniform, a multitude of images can beacquired to improve the statistics of a single measurement.

Figure 16 demonstrates the range of moduli that can be assessed using the wrin-kling metrology, as well as the precision of these measurements. Figure 16a showsmoduli data collected along a thickness gradient library of polystyrene [PS], illus-trating the ability of this technique to measure the modulus of glassy polymer films(E ≈ 1–5GPa), as well as its use in a combinatorial workflow. In the example shownin Fig. 16a, the average value for the modulus was 3.4GPa± 0.1GPa, in excellentagreement with reported bulk values for PS measured via conventional techniquessuch as tensile testing [73]. The surface wrinkling metrology can also measure softmaterials, such as poly(styrene–isoprene–styrene) [P(S–I–S)] block copolymers,that display moduli in the MPa range (Fig. 16b). Our surface wrinkling metrol-ogy can easily discriminate materials having less than a 5% difference in moduli,which demonstrates the unique power of this metrology for thin polymer film re-search. This point is illustrated in Fig. 16c, where the modulus of ultrathin PS filmsis shown to decrease sharply when the film thickness decreases below ≈ 40nm [74].

Fig. 16 Representative data from the surface wrinkling metrology, demonstrating the unprece-dented range of moduli and the precision that the methodology unlocks: a modulus of a thicknessgradient library of PS (reproduced with permission from [72]); b modulus as a function of compo-sition for P(S–I–S) triblock copolymer blends; c modulus as a function of thickness for ultrathin PSfilms (reproduced with permission from [74]). The lines are meant to guide the eye and the errorbars represent one standard deviation of the data, which is taken as the experimental uncertaintyof the measurement


Our metrology has also been applied to particularly challenging systems, such aslayer-by-layer assemblies [75–77], polymer brushes [78], and even single carbonnanotubes [79], with remarkable results.

The mechanics of surface wrinkling necessitate that there be a reasonable mod-ulus difference between the film of interest and the soft substrate (Es ≈ 2MPa forPDMS). In order to probe softer materials, the wrinkling metrology can be inverted,such that a sensor film of known modulus is adhered to a soft substrate of unknownmodulus. Rearrangement of (1) leads to the following expression for the modulusof the soft, elastic substrate:

Es =Ef

3

(λ

2πhf

)−3

(2)

Here, the unknown to be determined is the modulus of the soft elastic substrate, Es.The thickness of the sensor film is chosen such that the wavelength of the wrin-kling instability can again be measured by small angle light scattering (SALS),thus enabling high-throughput measurement of the substrate modulus. Experimen-tal validation of this approach was conducted using a series of model crosslinkedPDMS elastomers [80]. To extend the applicability of the buckling metrology aswell as demonstrate its versatility, we investigated its use for determining the modu-lus of commonly used and commercially relevant poly(HEMA) hydrogels, whichare widely used in the fields of contact lenses and biomaterials. Using this in-verted geometry, we measured moduli of hydrated elastomers between 0.21MPa <Es < 2.6MPa, greatly extending the demonstrated range of moduli that this surfacewrinkling metrology can probe. Substrates containing either discrete or continuousgradients in modulus (via gradients in composition or crosslink density, for exam-ple) can be easily integrated into this measurement workflow.

4.2 Adhesion Testing

The ability to control and tailor the adhesion at various interfaces plays a criticalrole in numerous technologies including electronic packaging, coatings and paints,biomedical implants, and pressure sensitive adhesives (PSAs). Previous researchhas shown that polymer interface formation and failure is dependent upon a range ofmaterial, processing and testing parameters. Current (traditional) approaches to thecharacterization of adhesion have focused on isolating a single adhesion-controllingparameter and correlating the changes in adhesion with corresponding changes inthat single parameter. However, these types of approaches are time-consuming, dis-crete, and do not allow interplay between variables to be investigated. Indeed, onemajor challenge in this field is the efficient exploration of this large parameter spacein order to develop an understanding of the fundamental driving forces for devel-opment of adhesive strength at polymer/polymer, polymer/metal, polymer/ceramic,and polymer/biomaterial interfaces. The ability to conduct highly parallel tests and


employ multivariant libraries is an essential step toward rapidly and efficientlyidentifying structure-property relationships critical for tuning adhesive performance[81, 82]. At NIST, we developed several combinatorial and high-throughput mea-surement platforms for probing both the fundamental origins of adhesion (e.g.,interfacial interactions) as well as practical aspects of adhesion in soft materials(e.g., PSAs) as well as glassy materials and thermosets (e.g., epoxies).

4.2.1 Viscoelastic Materials: Peel Tests

The peel test is one of the most common techniques to assess the adhesive prop-erties of PSAs. As the demand increases for combinatorial tools to test materialperformance rapidly, applying combinatorial and/or high-throughput approaches tothe peel test could yield valuable insight into PSA structure-property relationshipsas well as open the door to a larger parameter space that can be rapidly and ef-ficiently explored [83, 84]. However, there are considerable technical challengespresented by adapting conventional peel tests to include combinatorial or high-throughput concepts. To illustrate some of these challenges, we consider a simpleexample: measuring the adhesive strength of a commercial PSA tape adhered toa glass substrate possessing a surface energy gradient (Δγ) along the peel direction(see Fig. 17a). The peel force (F) is averaged across the peel width (b), thus severelylimiting the ability to apply an orthogonal gradient in this geometry. However, bycombining both a gradient that increases in surface energy (+Δγ) and one that de-creases in surface energy (−Δγ) relative to the peel direction, we can probe the effectthat the gradient itself has on the peeling process (e.g., crack acceleration or decel-eration, stick-slip, etc). As shown in Fig. 17b, the resulting peel data correlate wellwith the changing surface energy of the substrate. While this example demonstrates

Fig. 17 a Schematic of peeling of an adhesive tape off a surface energy gradient. b Peel data(F/b) as a function of distance (d) along the surface energy gradient. As a reference, a traditional(non-gradient) peel test was also performed on the low surface energy substrate


the utility and value of combinatorial approaches as applied to peel tests, it alsohighlights a drawback: the lack of ample statistical information that is inherent inthis type of measurement. For example, a conventional peel test conducted underconstant conditions results in a fluctuating force to be averaged. Applying a contin-uous gradient of sample properties or test conditions in the peel direction impliesthat each data point (force) corresponds to a single point in parameter space, thusprohibiting the average force to be calculated for a given condition. To address thisissue, we developed [83] a simple statistical treatment that allows a relationshipbetween the uncertainty of the force and the domain size to be established. Thisstatistical tool enables one to define the gradient step size (discrete gradients) orgradient steepness (continuous gradients) such that sound statistical information canbe obtained and measurement uncertainties can be defined.

4.2.2 Viscoelastic Materials: Probe Tack Tests

Another common method for measuring adhesion strength in soft adhesives is theprobe tack test, which involves bringing a rigid probe of known geometry into andout of contact with a flat adhesive layer while recording the applied displacementand resulting force throughout the cycle. In most cases, the probe geometry is ei-ther a cylindrical punch or a hemispherical lens. Since one needs to measure bothforce and displacement, it is difficult to design a parallel screening approach to thismeasurement platform. Therefore, the focus has been primarily on developing ap-propriate combinatorial libraries that span the parameter space of interest, whileemploying single point measurements of probe tack [85, 86]. At NIST we usedthis approach to examine the effect of temperature on the adhesive properties ofmodel PSAs (see Fig. 18). In this study, a temperature gradient is established acrossa transparent sapphire window that is coated with a soft adhesive. The transparentsubstrate allowed us to image simultaneously the contact zone from below the sam-ple, yielding valuable information about the failure mechanisms of the adhesive.Although tack tests were conducted in a serial manner across the temperature

Fig. 18 a Schematic of probe tack measurements of a thin adhesive film along a temperaturegradient. b Compilation of probe tack data during loading and unloading cycles for different tem-peratures. c Total adhesion energy, calculated from the area under the load-displacement curveshown in b divided by maximum contact area, as a function of temperature. The error bars rep-resent one standard deviation of the data, which is taken as the experimental uncertainty of themeasurement. (Reproduced with permission from [86])


gradient, our design yielded a dramatic decrease in the total measurement timeto adequately survey the entire temperature range studied. For example, adhesionmeasurements at ten different temperatures by conventional tack tests would take330 min [(30 min equilibration at each temperature +3min for tack test) ×10 dif-ferent temperatures], while adhesion measurements using the temperature gradienttack apparatus would only take 60 min [30 min equilibration on temperature gradi-ent + (3 min for tack test ×10 different temperatures)]. Thus, by incorporating atemperature gradient stage, we realized more than a fivefold increase in measure-ment throughput. We believe this new high-throughput technique has considerableanalytical utility because several critical pieces of information can be acquired si-multaneously and more efficiently, thus reducing experimental uncertainties andoverall measurement time. In this initial study, we conducted only 1D in situ tackmeasurements, opting to use the second dimension to conduct multiple identicaltests for statistical purposes, but more advanced applications are possible. For exam-ple, combinatorial aging tests and kinetic studies of epoxy curing can be examinedusing this instrument. By introducing another parameter such as aging or curingtime, 2D libraries (e.g., time and temperature) can be easily generated and screenedby probe-type tack measurements.

4.2.3 Glassy Materials: Edge-Delamination Tests

Combinatorial and high-throughput measurements of adhesion in rigid coatings andfilms demand quite different approaches than soft adhesives such as PSAs [81, 87,88]. For example, contact methods such as the probe tack test are not suitable formeasuring the interfacial strength of a coating adhered to a substrate. To this end,NIST researchers adapted the edge-delamination test [89] to evaluate the adhesionof combinatorial coating libraries to various substrates. In the edge-delaminationtest, thermal stress arising from cooling of a film/substrate system can propagatean initial crack along the film–substrate interface. Debonding occurs at a criticaltemperature (Tc) due to the stress concentration near the crack tip. The critical stress(σc) necessary to debond the film can be calculated using (3):

σc =Ef

1− v(αs −αf)(Tc −Tref) (3)

where E is the elastic modulus, α is the coefficient of thermal expansion, and Tref

is a reference temperature where the film is assumed to be in a stress-free state andis usually chose to be the Tg of the film. The subscripts “f” and “s” denote the filmand substrate, respectively. If the failure is assumed to take place in the film verynear the interface (cohesive failure), the thermal stress at the critical temperature fordebonding can be related to the fracture energy (KIC) of the film:

KIC = σ0

√hf

2(4)

Thus, KIC can be used as an accurate descriptor of the interfacial strength in afilm/substrate system.


Fig. 19 a Schematic of the combinatorial edge-delamination test for probing interfacial adhesionstrength. The gradient specimen is diced in order to mechanically isolate individual cells on thearray as well as to provide an initial interfacial crack. b An optical micrograph of an individualelement of the array illustrating the width and sharpness of the diced area. c Image of a gradientspecimen after failure showing a distinct transition from bonded to debonded areas of the specimen.(Reproduced with permission from [91])

In order to adapt the edge-delamination test to combinatorial workflows, we firstdeveloped the framework of the metrology; this included theory, experimental de-sign, stress analysis and simulation of the approach [90]. We could then employcombinatorial libraries that incorporate one or more adhesion-controlling parame-ters into the edge-delamination experimental design. For example, a film is coatedonto a rigid substrate in such a way that the film has a gradient (e.g., thickness, sur-face energy, composition) in one direction, which is then subdivided into an arrayof separate edge-delamination samples, as depicted in Fig. 19a. This dicing processserves to generate a pre-crack at the film/substrate interface as well as to isolatemechanically each cell of the array from neighboring cells. A second orthogonalgradient could also be incorporated into the experimental design. The specimen isthen slowly cooled and debonding events are observed for those sample cells hav-ing stresses greater than a critical value. These stresses depend on the combinationof local temperature and film thickness. A map of failures can be constructed as afunction of each unique combination of variable one and variable two (Fig. 19c).Subsequently, the interfacial strength between the film coating and the substrate canbe deduced from the failure map using (4). We have demonstrated this methodologyby probing the adhesion of thin PMMA films to a silicon substrate possessing a gra-dient in surface energy [91]. In that particular study, an epoxy stress-generating layerwas coated directly on the PMMA films; an orthogonal thickness gradient of epoxywas applied to generate a gradient in the stress profile. We have also employed com-positional gradients in epoxy films [92], as well as gradients in temperature (bothcuring temperature and quench temperature).

4.2.4 Elastic Materials: JKR Adhesion Tests

Oftentimes, in order to understand adhesion at the macroscale, we need to under-stand first the fundamental molecular interactions at interfaces. With this in mind,


Fig. 20 a Test geometry for a single lens JKR test. The applied load (P), displacement (δ ), andcontact radius (a) are measured during a complete loading/unloading cycle. b Multilens array usedfor combinatorial adhesion measurements across a sample library, which can be combined withautomated image analysis of the contact area across multiple lenses of the array to yield high-throughput measurements of adhesion across combinatorial libraries. c Representative data for Gfor both loading and unloading segments of PDMS/PDMS contact. (b reproduced with permissionfrom [95])

we developed a measurement platform to study quantitatively adhesion using theJohnson, Kendall and Roberts (JKR) model [93]. In this test, a hemispherical lensof one material is brought into and out of contact with a complementary substratewhile measuring the applied load, displacement, and contact area between the lensand substrate (see Fig. 20a). The energy release rate (G) represents the amount ofenergy required to change the contact area a unit amount, or more simply the ad-ditional energy required to drive the separation between the two surfaces, and isgiven by

G =(P′ −P)2

8πE∗a3 (5)

where P′ = 4E∗a3

3R is the Hertzian contact load (no adhesion), P is the applied load,E∗ is the system modulus, and a is the contact radius. For a JKR test, the energyrequired to increase surface area during the loading curve is bounded by the ther-modynamic work of adhesion (W ), while the unloading segment provides a measureof the adhesion hysteresis (GHYS), which reflects specific adhesion interactions thatdevelop while the lens is in contact with the substrate.

To facilitate high-throughput measurements on combinatorial libraries, we de-veloped means to introduce a planar array of hemispherical lenses (see Fig. 20b)into contact with a substrate possessing a gradient in material properties or envi-ronmental parameters along one or both axes of the array [94, 95]. Conversely, thehemispherical lens array could embody one of the property gradients, such as a gra-dient in surface energy, or the lens array itself could be comprised of a materialgradient, such as composition of the lenses within in the array. If two orthogonalgradients are placed on the array, then each lens contact point yields a measure-ment of adhesion for a unique combination of parameters. The challenge of using


a multilens array lies in the inability to measure load on each individual lens of thearray. In this case, the displacement (δ ) of the lens array (not the load (P), as in (5))is used to calculate G:

G =E∗ (δ ′ − δ)2

2πa(6)

where δ ′ = a2

R is the Hertzian contact displacement.We have employed microlens arrays that contain 100–1,600 individual lenses

per cm2, depending on the sample size and steepness of the gradient under study.We have analyzed the effect of surface energy, crosslink density, and contact timeusing our multilens measurement approach. By integrating a temperature gradientinto the instrument design, we can also measure temperature-dependent adhesiveproperties as a function of multiple variables. For example, in one study we mea-sured the self-adhesion and fracture of polystyrene thin films using orthogonalgradients in temperature and film thickness [94]. This methodology is a powerfultool for investigating the effects of multivariable environments (e.g., surface en-ergy, surface roughness, composition, and processing) on polymer adhesion. We arecurrently pursuing methods for functionalizing the PDMS lens array with differentchemistries in a graded manner, such as layer-by-layer deposition of polyelec-trolytes [96] and growth of polymer brushes [78], in ways that express the chemicaldiversity inherent in many interfaces.

5 High-Throughput Materials Synthesis and SolutionCharacterization: Microscale Approaches to PolymerLibrary Fabrication in Fluids

One of the major advantages of the techniques described above is the ability to mea-sure materials properties with significantly reduced quantities of sample. There arefew robust synthetic techniques, however, that can produce only the quantities ofpolymer necessary for these types of measurements. The development of microflu-idic devices presented an appealing technology for adaptation to addressing thisproblem by producing devices that could carry out polymer synthesis in microlitervolumes with cheap and flexible reactors directly interfaced with high-throughputmeasurement methods.

Additional potential advantages associated with performing chemical reactionsin the confined space of a microfluidic channel include improved heat transfer, uni-form mixing profiles, faster variable changes and improved safety [97]. Combinedwith a high-throughput characterization strategy this extends the versatility of a mi-crofluidic R&D platform to studying routes for improving control of molar mass,polydispersity, architecture and composition. Previous studies demonstrate that im-proved control of polymer products was obtained when reactions were carried outin a microfluidic or microreactor environment [98].


The NIST program had a two-pronged approach: (1) use microfluidic devices tomanipulate the stoichiometry and other conditions of reactions to produce contin-uous gradients in polymer chain structures, and (2) exploit the small volumes usedin microfluidic channels to carry out measurements on polymer solutions that tradi-tionally require significantly larger volumes or longer times to measure. Examplesthat will be presented in this section include the preparation of continuous gradi-ents of molecular weight using ATRP, dynamic light scattering (DLS) on a chip andhigh-throughput interfacial tension (IFT) measurements.

5.1 Controlled Polymer Synthesis in Microchannels

With the development of controlled radical polymerization, synthetic polymerchemists dramatically expanded the range of materials that could be controlled onthe molecular scale. Controlled and high-throughput techniques are essential to thesystematic survey of this vast parameter space. NIST chose to pursue the use of atechnique with flexibility, ease of use, and sample volumes similar to those usedin the other library design and measurement methods described above. Channelsfabricated in both polymer/glass (Fig. 21a) [99, 100] and metal devices [101] wereused depending on the conditions necessary for the reactions. Earlier polymericdevices were replaced by metallic devices when higher temperatures and longerreaction times were required.

The first reactions were carried out at room temperature in devices fabricatedfrom a thiolene resin cured between two glass slides. 2-Hydroxypropyl methacrylate(HPMA) was polymerized by ATRP, and reaction kinetics similar to those obtainedin a traditional batch reaction were obtained by adjusting the total flow rate of thefluid through the channel and treating the residence time in the channel as the reac-tion time (Fig. 21b,c) [102].

The correlation of residence time to reaction time is critical in the ability to treatthe volume of the channel as a continuous gradient in molecular mass. ATRP isparticularly well-suited to this type of device because the reaction can be initi-ated at a fixed mixing element at the head of the channel where a catalyst andinitiator can be brought together. By replacing a small molecule initiator with apolymer chain capable of being reinitiated, copolymers could be prepared. Thiswas done in the thiolene/glass devices with a poly(n-butyl methacrylate) block[103].

In order to access a wider variety of monomers, higher temperatures were neces-sary. Using an aluminum channel capped on one wall with a Kapton film, styrene,as well as several acrylates and methacrylates, were polymerized. Furthermore,block copolymers were also prepared from these more widely used polymers, andthe devices were integrated with characterization techniques as described below[104]. Similar devices have been used to carry out anionic polymerizations as well[105].


a

b

c

Initiatorin solvent

Catalystin monomer

Reaction Time (h)

Pum

ping Rate (mL/h)

Mixing element (initiation point)

2.5

2000

4000

6000

8000

10000

12000

400

200

600

800

1000

1200

1400

2.5

2.0

2

2

1.5

1.5

1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

1

1

0.5

0.50.00

0 2 4

Time (hr)

Flask

CRP Chip

PDI for flask

PDI for CRP Chip

6 8 10 120

0

Mw

/Mn

Mn

In (

[M] o/

[M])

Fig. 21 Representative microfluidic device and resulting data from ATRP on a chip: a image ofa microfluidic device (dimensions 25mm×75mm) fabricated from UV curable thiolene resin be-tween two glass slides; b reaction data for ATRP of HPMA synthesized on a chip showing thecorrelation of flow rate (or residence time) to reaction time and resulting conversion of monomer(M) to polymer (ln([M]o/[M]); c comparison of number average molecular mass (Mn) and poly-dispersity for n-butyl acrylate prepared in a traditional round bottom flask (‘Flask’) and on a chip(‘CRP Chip’). (Reproduced with permission from [102])

5.2 Characterization of Interfacially-Active Polymersin Microchannels

In order to obtain the advantage of other high-throughput and combinatorial tech-niques in microfluidic reactors, it is critical that other processing and measurement


20

15

C

10

DH, n

m

5

0 0

20

40

60

Intensity, kcps

80

100

120

24 26 28 30 32 34 36 38 40

Temperature, °C42

Fig. 22 Images and data representing development and application of DLS on a chip: a one it-eration in the design of a microfluidic DLS fabricated from aluminum with the surface anodizedblack to reduce surface reflections; b image of a microfluidic chip that integrates polymer synthe-sis with DLS. The machined channels have been covered by a Kapton sheet fixed with adhesive;c data for temperature depended micelle formation of polyethylene oxide–polypropylene oxide–polyethylene oxide triblock copolymer (Pluronic P85) at 2% by volume in water. (Derived from[106] with permission)

tools be integrated into the same platform. The NIST team designed a dynamiclight scattering (DLS) instrument on a chip by anodizing the inner walls ofan aluminum channel to minimize reflections and plumbing the channel withfiber optics which both deliver an incident beam and detect off-axis scattering(Fig. 22a). After several iterations of design [106], the DLS was applied to detectthe formation of micelles in block copolymers as a function of cosolvent fraction(polystyrene/polyisoprene diblock copolymers in hexanes/toluene) or temperature(polyethylene oxide/polypropylene oxide triblock copolymers in water; Fig. 22c).

This new DLS tool was integrated into the microreactors by placing it at the endof a chip that synthesized amphiphilic block copolymers from methyl methacrylateand either lauryl methacrylate or octadecyl methacrylate (Fig. 22b). The reaction


mixture was then diluted with a selective solvent while still on the device, whichboth terminated the polymerization and induced micelle or other structure forma-tion in the fluid. The structured samples then flowed through the DLS chamber forcharacterization of size [101]. The total sample volume in the DLS measurementchamber was only 4μL, and the device enabled comparatively simple alignmentprocedures while reducing multiple scattering.

The integrated DLS device provides an example of a measurement tool tailored tonano-scale structure determination in fluids, e.g., polymers induced to form specificassemblies in selective solvents. There is, however, a critical need to understand thebehavior of polymers and other interfacial modifiers at the interface of immisciblefluids, such as surfactants in oil-water mixtures. Typical measurement methods usedto determine the interfacial tension in such mixtures tend to be time-consumingand had been described as a major barrier to systematic surveys of variable spacein libraries of interfacial modifiers. Critical information relating to the behavior ofsuch mixtures, for example, in the effective removal of soil from clothing, wouldbe available simply by measuring interfacial tension (IFT) for immiscible solutionswith different droplet sizes, a variable not accessible by drop-volume or pendantdrop techniques [107].

Through many iterations of design, NIST scientists developed a microfluidic chipthat addressed this challenge, by forming droplets of immiscible fluids in a contin-uous flow stream, while systematically varying conditions that would influence theinterfacial tension between the two fluids. The device consisted of three basic ele-ments: drop formation, mixing and drop adsorption, and drop deformation, followedby detection of the drop relaxation using a charge-coupled device (CCD) camera.A variety of mixtures were characterized with these techniques, including the silox-ane interface with water, air, ethylene glycol and glycerol, with and without addedsurfactants (Fig. 23a) [108].

Challenges associated with the proper design of the instrument on a chip includedconsideration and elimination of confinement effects, depletion issues when the sur-factant concentration was determined by the size of the droplet (i.e., the dropletphase contained the surfactant additive) and full automation of fluid controls, im-age capture and data analysis. Ultimately, however, the device was demonstratedto measure both equilibrium and dynamic IFT in a fraction of the time necessaryby conventional techniques and at a length-scale of greater relevance to the appli-cations of interest. The rapid scanning of composition variation was demonstratedby measuring the IFT of water/ethylene glycol mixtures in polydimethylsiloxane oil(Fig. 23b) [109].

6 Conclusions

We provided an overview of combinatorial and high-throughput methods researchat NIST, with a focus on tools and application examples that are useful for the exam-ination of polymer surfaces, interfaces and thin films. An examination of this body


a

b

σ (m

N/m

)

1a

1 cm

1b

3a

3b

2

0

10

0.0 0.2 0.4

φEG

0.6 0.8 1.0

20

30

40

50

4

Fig. 23 a Image of a microfluidic chip used for IFT measurements filled with liquid dye to il-luminate channels. To perform the measurement, drops are injected (fluid 1a and b) are injectedinto an immiscible stream (2). Additional immiscible matrix is added (3a and 3b) conveying thedrops into channel 4 for analysis and measurement. Constrictions in channel 4 accelerate/stretchthe drops. Multiple constrictions enable measurement at different interface age. The channel ge-ometry is shown schematically in the inset (from [108]). b Interfacial tension (σ ) of water/ethyleneglycol mixtures (binary drops) in PDMS oil, as a function of composition (φ ). (Reproduced withpermission from [109])

of work illustrates two key points that are worth discussing in conclusion. First, itis clear that gradient and microfluidic methods are powerful tools for polymer re-search, and this is not only because these techniques can be more rapid (although,this is one great advantage). A more important aspect, especially for emerging poly-mer technologies, is that these techniques allow scientists and engineers to approachcomplex materials systems in ways that are impossible via traditional techniques.As illustrated in many of the examples discussed above, a library approach enablesthe researcher to view, often in a single specimen, an entire space of structures,behaviors and their response to influencing variables. These “big picture” snap-shots can be invaluable in initiating and building a comprehensive understanding


of complex systems and the factors that govern them. In addition, gradient librariesenable the researcher to “zoom in” on specific parameter subsets to achieve moredetailed analyses; this unique capability facilitates a more rigorous examination ofthe structure-processing-performance interrelationships that are key to materials sci-ence and engineering.

The second point is that, quite often, library approaches can provide significantadvantages even if they are not integrated into extensive workflow infrastructuresthat are often associated with combinatorial methods. Indeed, gradient libraries andmicrofluidic library synthesis can provide extensive benefits both in their own rightand in combination with modest automated analysis. The “self-reporting” aspect ofgradient libraries is one example of this, as is the ability of continuous microreac-tors to create systematically changing families of polymer specimens while usingminiscule amounts of reactant. Moreover, by pairing libraries with high-throughputmeasurement platforms researchers have access to an unparalleled, and rapid, abilityto determine the factors that govern and optimize materials performance. Examplesof this can be seen in our discussion of high-throughput adhesion and mechanicalproperties tests, which are fueled by appropriately fabricated library specimens. Inaddition, as we discussed, microfluidic technologies are quite powerful in this re-spect, since they can integrate library fabrication and high-throughput analysis offluid specimens on a single device.

In each of these “take home” messages, the benefits of combinatorial and high-throughput approaches are gained by thinking beyond the single sample paradigm,by being aware of the opportunities afforded by developing and exploiting thesetools, and by applying them with the wisdom that has always characterized success-ful polymers science.

References

1. Webster DC (2005) Radical change in research and development: the shift from conventionalmethods to high-throughput methods. JCT Coat Technol 2:24–29

2. Hanak JJ (1970) Multiple-sample-concept in materials research – synthesis, compositionalanalysis and testing of entire multicomponent systems. J Mater Sci 5:964

3. Meredith JC, Karim A, Amis EJ (2000) High-throughput measurement of polymer blendphase behavior. Macromolecules 33:5760–5762

4. Meredith JC, Karim A, Amis EJ (2002) Combinatorial methods for investigations in polymermaterials science. MRS Bull 27:330–335

5. Stafford CM, Roskov KE, Epps TH, Fasolka MJ (2006) Generating thickness gradients ofthin polymer films via flow coating. Rev Sci Instrum 77:023908

6. Zhang X, Berry BC, Yager KG, Kim S, Jones RL, Satija S, Pickel DL, Douglas JF, Karim A(2008) Surface morphology diagram for cylinder-forming block copolymer thin films. ACSNano 2:2331–2341

7. de Gans BJ, Wijnans S, Woutes D, Schubert US (2005) Sector spin coating for fast preparationof polymer libraries. J Comb Chem 7:952–957

8. Cull TR, Goulding MJ, Bradley M (2007) Liquid crystal libraries-ink-jet formulation andhigh-throughput analysis. Adv Mater 19:2355–2359

9. Meier MAR, Schubert US (2005) Integration of MALDI-TOFMS as high-throughput screen-ing tool into the workflow of combinatorial polymer research. Rev Sci Instrum 76:062211


10. Yoshioka Y, Calvert PD, Jabbour GE (2005) Simple modification of sheet resistivity of con-ducting polymeric anodes via combinatorial ink-jet printing techniques. Macromol RapidCommun 26:238–246

11. de Gans BJ, Kazancioglu E, Meyer W, Schubert US (2004) Ink-jet printing polymers andpolymer libraries using micropipettes. Macromol Rapid Commun 25:292–296

12. Koner L, Kofler W (1949) A heating bed for rapid determination of the melting point.Mikrochem Mikrohim Acta 34:374–381

13. Sehgal A, Karim A, Stafford CF, Fasolka MJ (2003) Techniques for combinatorial and high-throughput microscopy: part 1: gradient specimen fabrication for polymer thin films research.Micros Today 11:26–29

14. Beers KL, Douglas JF, Amis EJ, Karim A (2003) Combinatorial measurements of crys-tallization growth rate and morphology in thin films of isotactic polystyrene. Langmuir19:3935–3940

15. Lucas LA, DeLongchamp DM, Vogel BM, Lin EK, Fasolka MJ, Fischer DA, McCulloch I,Heeney M, Jabbour GE (2007) Combinatorial screening of the effect of temperature on themicrostructure and mobility of a high performance polythiophene semiconductor. Appl PhysLett 90:012112

16. Eidelman N, Raghavan D, Forster AM, Amis EJ, Karim A (2004) Combinatorial approach tocharacterizing epoxy curing. Macromol Rapid Commun 25:259–263

17. Roberson SV, Faheya AJ, Sehgal A, Karim A (2002) Multifunctional ToF-SIMS: combinato-rial mapping of gradient energy substrates. Appl Surf Sci 200:150–164

18. Smith AP, Sehgal A, Douglas JF, Karim A, Amis EJ (2003) Combinatorial mapping of surfaceenergy effects on diblock copolymer thin film ordering. Macromol Rapid Commun 24:131–135

19. Berry BC, Stafford CM, Pandya M, Lucas LA, Karim A, Fasolka MJ (2007) Versatile plat-form for creating gradient combinatorial libraries via modulated light exposure. Rev SciInstrum 78:072202

20. Elwing H, Welin S, Askendal A, Nilsson U, LundstrÖm I (1987) A wettability gradientmethod for studies of macromolecular interactions at the liquid/solid interface. J Colloid In-terface Sci 119:203–210

21. Chaudhury MK, Whitesides GM (1992) How to make water run uphill. Science 256:-1539–1541

22. Genzer J (2005) Templating surfaces with gradient assemblies. J Adhes 81:417–43523. Zhao H, Beysens D (1995) From droplet growth to film growth on a heterogeneous surface –

condensation associated with a wettability gradient. Langmuir 11:627–63424. Liedberg B, Tengvall P (1995) Molecular gradients of omega-substituted alkanethiols on

gold – preparation and characterization. Langmuir 11:3821–382725. Liedberg B, Wirde M, Tao YT, Tengvall P, Gelius U (1997) Molecular gradients of omega-

substituted alkanethiols on gold studied by X-ray photoelectron spectroscopy. Langmuir13:5329–5334

26. Morgenthaler S, Lee S, Zurcher S, Spencer ND (2003) A simple, reproducible approach tothe preparation of surface-chemical gradients. Langmuir 19:10459–10462

27. Morgenthaler SM, Lee S, Spencer ND (2006) Submicrometer structure of surface-chemicalgradients prepared by a two-step immersion method. Langmuir 22:2706–2711

28. Ashley KM, Raghavan D, Douglas JF, Karim A (2005) Wetting-dewetting transition line inthin polymer films. Langmuir 21:9518–9523

29. Fasolka MJ, Mayes AM (2001) Block copolymer thin films: physics and applications. AnnuRev Mater Res 31:323–355

30. Smith AP, Douglas JF, Meredith JC, Amis EJ, Karim A (2001) Combinatorial study of surfacepattern formation in thin block copolymer films. Phys Rev Lett 87:015503

31. Ludwigs S, Schmidt K, Stafford CM, Amis EJ, Fasolka MJ, Karim A, Magerle R, KrauschG (2005) Combinatorial mapping of the phase behavior of ABC triblock terpolymers in thinfilms: Experiments. Macromolecules 38:1850–1858


32. Epps TH, Delongchamp DM, Fasolka MJ, Fischer DA, Jablonski EL (2007) Substrate surfaceenergy dependent morphology and dewetting in an ABC triblock copolymer film. Langmuir23:3355–3362

33. Roskov KE, Epps TH, Berry BC, Hudson SD, Tureau MS, Fasolka MJ (2008) Preparationof combinatorial arrays of polymer thin films for transmission electron microscopy analysis.J Comb Chem 10:966–973

34. Julthongpiput D, Fasolka MJ, Zhang WH, Nguyen T, Amis EJ (2005) Gradient chemicalmicropatterns: a reference substrate for surface nanometrology. Nano Lett 5:1535–1540

35. Julthongpiput D, Zhang W, Douglas JF, Karim A, Fasolka MJ (2007) Pattern-directed toisotropic dewetting transition in polymer films on micropatterned surfaces with differentialsurface energy contrast. Soft Matter 3:613–618

36. Gallant ND, Lavery KA, Amis EJ, Becker ML (2007) Universal chemistry for “Click”chemistry biofunctionalization. Adv Mater 19:072207

37. Kolb HC, Finn MG, Sharpless BK (2001) Click chemistry: diverse chemical function from afew good reactions. Angew Chem Int Ed 40:2004–2021

38. Lin-Gibson S, Landis FA, Drzal PL (2006) Combinatorial investigation of the structure-properties characterization of photopolymerized dimethacrylate networks. Biomaterials27:1711–1717

39. Lin NJ, Drzal PL, Lin-Gibson S (2007) Two-dimensional gradient platforms for rapid as-sessment of dental polymers: a chemical, mechanical and biological evaluation. Dent Mater23:1211–1220

40. Johnson PM, Stansbury JW, Bowman CN (2007) Photopolyrner kinetics using light intensitygradients in high-throughput conversion analysis. Polymer 48:6319–6324

41. Johnson PM, Stansbury JW, Bowman CN (2008) High-throughput kinetic analysis of acry-late and thiol-ene photopolymerization using temperature and temperature and exposure timegradients. J Polym Sci Part A Polym Chem 46:1502–1509

42. Johnson PM, Stansbury JW, Bowman CN (2008) Kinetic modeling of a comonomer photo-polymerization system using high-throughput conversion data. Macromolecules 41:230–237

43. Matyjaszewski K, Miller PJ, Shukla N, Immaraporn B, Gelman A, Luokala BB, Siclovan TM,Kickelbick G, Vallant T, Hoffmann H, Pakula T (1999) Polymers at interfaces: using atomtransfer radical polymerization in the controlled growth of homopolymers and block copoly-mers from silicon surfaces in the absence of untethered sacrificial initiator Macromolecules32:8716–8724

44. Husseman M, Malmstrom EE, McNamara M, Mate M, Mecerreyes D, Benoit DG, HedrickJL, Mansky P, Huang E, Russell TP, Hawker CJ (1999) Controlled synthesis of polymerbrushes by “Living” free radical polymerization techniques. Macromolecules 32:1424–1431

45. Bhat RR, Tomlinson MR, Genzer J (2005) Orthogonal surface-grafted polymer gradients: aversatile combinatorial platform. J Polym Sci Part B Polym Phys 43:3384–3394

46. Genzer J, Bhat RR (2008) Surface-bound soft matter gradients. Langmuir 24(6):2294–231747. Wu T, Efimenko K, Genzer J (2002) Combinatorial study of the mushroom-to-brush crossover

in surface anchored polyacrylamide. J Am Chem Soc 124:9394–939548. Wu T, Efimenko K, Vlcek P, Subr V, Genzer J (2003) Formation and properties of anchored

polymers with a gradual variation of grafting densities on flat substrates. Macromolecules36:2448–2453

49. Zhao B, Brittain WJ (2000) Polymer brushes: surface-immobilized macromolecules. ProgPolym Sci 25:677–710

50. Patten TE, Matyjaszewski K (1998) Atom transfer radical polymerization and the synthesisof polymeric materials. Adv Mater 10:901

51. Mendes PM (2008) Stimuli-responsive surfaces for bio-applications. Chem Soc Rev37:2512–2529

52. La WH, Wang RM, He YF, Zhang HF (2008) Preparation and application of smart coatings.Prog Polym Chem 20:351–361

53. Patten TE, Xia JH, Abernathy T, Matyjaszewski K (1996) Polymers with very low polydis-persities from atom transfer radical polymerization. Science 272:866–868


54. Prucker O, Ruhe J (1998) Synthesis of poly(styrene) monolayers attached to high surface areasilica gels through self-assembled monolayers of azo initiators. Macromolecules 31:592–601

55. Prucker O, Ruhe J (1998) Polymer layers through self-assembled monolayers of initiators.Langmuir 14:6893–6898

56. Xu C, Wu T, Drain CM, Batteas JD, Beers KL (2005) Microchannel confined surface-initiatedpolymerization. Macromolecules 38:6

57. Xu C, Wu T, Batteas JD, Drain CM, Beers KL, Fasolka MJ (2006) Surface-graftedblock copolymer gradients: effect of block length on solvent response. Appl Surf Sci252:2529–2534

58. Xu C, Wu T, Drain CM, Batteas JD, Fasolka MJ, Beers KL (2006) Effect of block lengthon solvent response of block copolymer brushes: combinatorial study with block copolymerbrush gradients. Macromolecules 39:3359–3364

59. Mei Y, Wu T, Xu C, Langenbach KJ, Elliott JT, Vogt BD, Beers KL, Amis EJ, Washburn NR(2005) Tuning cell adhesion on gradient poly(2-hydroxyethyl methacrylate)-grafted surfaces.Langmuir 21:12309–12314

60. Xu C, Barnes SE, Wu T, Fischer DA, DeLongchamp DM, Batteas JD, Beers KL (2006) Solu-tion and surface gradients via microfluidic confinement: fabrication of a statistical copolymerbrush composition gradient. Adv Mater 18:1427

61. Zhao B (2004) A combinatorial approach to study solvent-induced self-assembly of mixedpoly(methyl methacrylate)/polystyrene brushes on planar silica substrates: effect of relativegrafting density. Langmuir 20:11748–11755

62. Patton DL, Page KA, Xu C, Genson KL, Fasolka MJ, Beers KL (2007) Measurement ofreactivity ratios in surface-initiated radical copolymerization. Macromolecules 40:6017–6020

63. Zhao JC (2006) Combinatorial approaches as effective tools in the study of phase diagramsand composition-structure-property relationships. Prog Mater Sci 51:557–631

64. Karim A, Yurekli K, Meredith C, Amis EJ, Krishnamoorti R (2002) Combinatorial methodsfor polymer materials science: phase behavior of nanocomposite blend films. Polym Eng Sci42:1836–1840

65. Meredith JC, Sormana JL, Keselowsky BG, Garcia AJ, Tona A, Karim A, Amis EJ (2003)Combinatorial characterization of cell interactions with polymer surfaces. J Biomed MaterRes Part A 66A:483–490

66. Simon CG Jr, Stephens JS, Dorsey SM, Becker ML (2007) Fabrication of combinatorial poly-mer scaffold libraries. Rev Sci Instrum 78:0722071

67. Yang Y, Dorsey SM, Becker ML, Lin-Gibson S, Schumacher GE, Flaim GM, Kohn J, SimonCG Jr (2008) X-ray imaging optimization of 3D tissue engineering scaffolds via combinato-rial fabrication methods. Biomaterials 29:1901–1911

68. Kramer EJ (1983) Microscopic and molecular fundamentals of crazing. Adv Polym Sci52/53:1–56

69. Lauterwasser BD, Kramer EJ (1979) Microscopic mechanisms and mechanics of crazegrowth and fracture. Philos Mag A Phys Condens Matter Struct Defects Mech Prop39:469–495

70. Crosby AJ, Fasolka MJ, Beers KL (2004) High-throughput craze studies in gradient thin filmsusing ductile copper grids. Macromolecules 37:9968–9974

71. Stafford CM, Guo S, Harrison C, Chiang MYM (2005) Combinatorial and high-throughputmeasurements of the modulus of thin polymer films. Rev Sci Instrum 76:062207

72. Stafford CM, Harrison C, Beers KL, Karim A, Amis EJ, Vanlandingham MR, Kim HC,Volksen W, Miller RD, Simonyi EE (2004) A buckling-based metrology for measuring theelastic moduli of polymeric thin films. Nat Mater 3:545–550

73. Brandrup J, Immergut EH, Grulke EA (1999) Polymer handbook, 4th edn. Wiley, New York74. Stafford CM, Vogt BD, Harrison C, Julthongpiput D, Huang R (2006) Elastic Moduli of

ultrathin amorphous polymer films. Macromolecules 39:5095–509975. Nolte AJ, Cohen RE, Rubner MF (2006) A two-plate buckling technique for thin film mod-

ulus measurements: applications to polyelectrolyte multilayers. Macromolecules 39:4841–4847


76. Nolte AJ, Rubner MF, Cohen RE (2005) Determining the Young’s modulus of polyelec-trolyte multilayer films via stress-induced mechanical buckling instabilities. Macromolecules38:5367–5370

77. Nolte AJ, Treat ND, Cohen RE, Rubner MF (2008) Effect of relative humidity on the Young’smodulus of polyelectrolyte multilayer films and related nonionic polymers. Macromolecules41:5793–5798

78. Huang H, Chung JY, Nolte AJ, Stafford CM (2007) Characterizing polymer brushes via sur-face wrinkling. Chem Mater 19:6555–6560

79. Khang DY, Xiao JL, Kocabas C, MacLaren S, Banks T, Jiang HQ, Huang YYG, Rogers JA(2008) Molecular scale buckling mechanics on individual aligned single-wall carbon nan-otubes on elastomeric substrates. Nano Lett 8:124–130

80. Wilder EA, Guo S, Lin-Gibson S, Fasolka MJ, Stafford CM (2006) Measuring the modulusof soft polymer networks via a buckling-based metrology. Macromolecules 39:4138–4143

81. Chisholm B, Potyrailo R, Cawse J, Shaffer R, Brennan M, Molaison C, Whisenhunt D,Flanagan B, Olson D, Akhave J, Saunders D, Mehrabi A, Licon M (2002) The development ofcombinatorial chemistry methods for coating development – I. Overview of the experimentalfactory. Prog Org Coat 45:313–321

82. Crosby AJ (2003) Combinatorial characterization of polymer adhesion. J Mater Sci38:4439–4449

83. Chiche A, Zhang WH, Stafford CM, Karim A (2005) A new design for high-throughput peeltests: statistical analysis and example. Meas Sci Technol 16:183–190

84. McGuiggan PM, Chiche A, Filliben JJ, Phelan FR, Fasolka MJ, Yarusso DJ (2006) High-throughput peel measurement of a pressure-sensitive adhesive. Adhes Mag 13:32–39

85. Grunlan JC, Holguin DL, Chuang HK, Perez I, Chavira A, Quilatan R, Akhave J, Mehrabi AR(2004) Combinatorial development of pressure-sensitive adhesives. Macromol Rapid Com-mun 25(1):286–291

86. Moon SH, Chiche A, Forster AM, Zhang WH, Stafford CM (2005) Evaluation oftemperature-dependent adhesive performance via combinatorial probe tack measurements.Rev Sci Instrum 76:062210

87. Chisholm B, Potyrailo R, Shaffer R, Cawse J, Brennan M, Molaison C (2003) Combinatorialchemistry methods for coating development III. Development of a high-throughput screen-ing method for abrasion resistance: correlation with conventional methods and the effects ofabrasion mechanism. Prog Org Coat 47:112–119

88. Chisholm BJ, Potyrailo RA, Cawse JN, Shaffer RE, Brennan M, Molaison CA (2003) Com-binatorial chemistry methods for coating development V. The importance of understandingprocess capability. Prog Org Coat 47:120–127

89. Shaffer EO, McGarry FJ, Hoang L (1996) Designing reliable polymer coatings. Polym EngSci 36:2375–2381

90. Chiang MYM, Wu WL, He JM, Amis EJ (2003) Combinatorial approach to the edge-delamination test for thin film reliability – concept and simulation. Thin Solid Films437:197–203

91. Chiang MYM, Song R, Crosby AJ, Karim A, Chiang CK, Amis EJ (2005) Combinatorialapproach to the edge-delamination test for thin film reliability – adaptability and variability.Thin Solid Films 476:379–385

92. Stafford CM, Kim JH, Kawaguchi D, Royston G, Chiang MYM (2006) Probing the interfa-cial adhesion strength in compositional libraries of epoxy films. Mater Res Soc Symp Proc894:129–137

93. Johnson KL, Kendall K, Roberts AD (1971) Surface energy and contact of elastic solids. ProcR Soc Lond A Math Phys Sci 324:301

94. Crosby AJ, Karim A, Amis EJ (2003) Combinatorial investigations of interfacial failure.J Polym Sci Part B Polym Phys 41:883–891

95. Forster AM, Zhang WH, Crosby AJ, Stafford CM (2005) A multilens measurement platformfor high-throughput adhesion measurements. Meas Sci Technol 16:81–89

96. Nolte AJ, Chung JY, Walker ML, Stafford CM (2009) In-situ adhesion measurements utilizinglayer-by-layer functionalized surfaces. ACS Appl Mater Interfaces 1:373–380


97. Hessel V, Serra C, Loewe GH, Hadziioannou G (2005) Polymerizations in micro-structuredreactors: overview. Chem Ing Tech 77:1693

98. Iwasaki T, Yoshida J (2005) Free radical polymerization in microreactors. Significant im-provement in molecular weight distribution control. Macromolecules 38:1159–1163

99. Harrison C, Cabral J, Stafford CM, Karim A, Amis EJ (2004) A rapid prototyping techniquefor the fabrication of solvent-resistant structures. J Micromech Microeng 14:153–158

100. Cygan ZT, Cabral JT, Beers KL, Amis EJ (2005) Microfluidic platform for the generation oforganic-phase microreactors. Langmuir 21:3629–3634

101. Chastek TQ, Iida K, Amis EJ, Fasolka MJ, Beers KL (2008) A microfluidic platform forintegrated synthesis and dynamic light scattering measurement of block copolymer micelles.Lab Chip 8:950

102. Wu T, Mei Y, Cabral JT, Xu C, Beers KL (2004) A new synthetic method for controlledpolymerization using a microfluidic system. J Am Chem Soc 126:7881

103. Wu T, Mei Y, Xu C, Byrd HCM, Beers KL (2005) Block copolymer PEO-b-PHPMA synthe-sis using controlled radical polymerization on a chip. Macromol Rapid Commun 26:1037

104. Chastek TQ, Iida K, Amis EJ, Fasolka MJ, Beers KL (2008) A microfluidic platform forintegrated synthesis and dynamic light scattering measurement of block copolymer micelles.Lab Chip 8:950–957

105. Iida K, Chastek TQ, Beers KL, Cavicchi KA, Chun J, Fasolka MJ (2009) Living anionicpolymerization using a microfluidic reactor. Lab Chip 9:339–345

106. Chastek TQ, Beers KL, Amis EJ (2007) Miniaturized dynamic light scattering instrumenta-tion for use in microfluidic applications. Rev Sci Instrum 78:072201

107. Jin F, Balasubramaniam R, Stebe KJ (2004) Surfactant adsorption to spherical particles: theintrinsic length-scale governing the shift from diffusion to kinetic-controlled mass transfer.J Adhes 80:773–796

108. Hudson SD, Cabral JT, Goodrum WJ, Beers KL, Amis EJ (2005) Microfluidic interfacialtensiometry. Appl Phys Lett 87:081905

109. Cabral JT, Hudson SD (2006) Microfluidic approach for rapid multicomponent interfacialtensiometry. Lab Chip 6:427–436


Published online: 26 January 2010

Polymer Informatics

Nico Adams

Abstract Polymers are arguably the most important set of materials in commonuse. The increasing adoption of both combinatorial as well as high-throughput ap-proaches, coupled with an increasing amount of interdisciplinarity, has wroughttremendous change in the field of polymer science. Yet the informatics tools re-quired to support and further enhance these changes are almost completely absent.In the first part of the chapter, a critical analysis of the challenges facing modernpolymer informatics is provided. It is argued, that most of the problems facing thefield today are rooted in the current scholarly communication process and the wayin which chemists and polymer scientists handle and publish data. Furthermore, thechapter reviews existing modes of representing and communicating polymer infor-mation and discusses the impact, which the emergence of semantic technologieswill have on the way in which scientific and polymer data is published and transmit-ted. In the second part, a review of the use of informatics tools for the prediction ofpolymer properties and in silico design of polymers is offered.

Keywords Information systems · Machine learning · Ontology · Polymer markuplanguage · Polymer informatics · QSPR · RDF · Semantic web

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1092 Polymer Information is Challenging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

2.1 The Representation of Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1132.2 Access to Polymer Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

N. Adams (�)Unilever Centre for Molecular Science Informatics, University Chemical Laboratory,University of Cambridge, Lensfield Road, Cambridge CB2 1EW, UKemail: [email protected]

108 N. Adams

3 Making Use of Polymer Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1303.1 Classification and Chemometrics Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1303.2 Property Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

4 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

Abbreviations

ACS American Chemical SocietyANN Artificial neural networkBPAC Bisphenol A PolycarbonateCAS Chemical Abstracts ServiceFTIR Fourier Transform Infrared SpectroscopyGREMAS Genealogical Retrieval by Magnetic Tape StorageHIM Hamiltonian Interaction ModelingHTE High Throughput ExperimentationIUPAC International Union of Pure and Applied ChemistryLCST Lower Critical Solution TemperatureLDPE Low Density PolyethyleneLLDPE Linear Low Density PolyethyleneOWL Web Ontology LanguagePCA Principal Component AnalysisPCR Principal Component RegressionPDF Portable document formatPDI Polydispersity IndexPET Poly(ethylene terephthalate)POLIDCASYR Polymer Documentation System of IDC with Inclusion of Ana-

lytical and Synthetic Concept RelationsPVA Poly(vinyl alcohol)QSPR Quantitative Structure Property RelationshipR2 Correlation coefficientR2

cv Cross-validated correlation coefficientRBF Radial Basis FunctionRDF Resource Description FrameworkRMS Root Mean Square ErrorSTM Scientific, technical, medicalTg Glass transition temperatureToF-SIMS Time-of-Flight Secondary Ion Mass SpectrometryTOSAR Topological Representation of Synthetic and Analytical Rela-

tions of ConceptsUCST Upper Critical Solution TemperatureUV UltravioletWWW World Wide WebXML eXtensible Markup Language

Polymer Informatics 109

1 Introduction

Synthetic polymers are arguably the most important class of materials of the modernage. While it is difficult to obtain precise numbers, market research suggests that theworld production of polymers in 2003 was between 150 and 200 million tons a year.Within this mix, polyethylene is probably the most important commodity polymerand accounts for approximately 50 million tons a year. The consumption of linearlow density polyethylene (LLDPE) in Russia alone has been predicted to increasefrom 90,000 metric tons in 2007 to 1,00,000 metric tons in 2008 and the percentageincrease in LLDPE use in China is in double-digit figures. The significant consump-tion is explained by the fact that polymers have found a wide area of applicationin virtually all areas of life, from simple packaging and building materials to moresophisticated uses in engineering applications requiring high-performance materi-als as well as in medicine (both as biomaterials [1–3] and in drug-[4–11] and genedelivery [12–19]), home and personal care applications [20, 21] and as plastic andprinted electronics [22–26]. The reason for the success of polymers as materials canbe found in the combination of a number of reasons: (1) commodity polymers arecheap to manufacture (the average cost of crude-derived commodity polymers is ap-proximately 1d/kg) [27], (2) polymers are largely non-toxic, (3) polymers are easyto process, (4) polymers are easy to adapt through post-processing and additives and(5) polymers are on the whole stable and relatively tough.

For the reasons outlined above, polymer science is very much in the mainstreamof both chemical and materials research and likely to remain there for the fore-seeable future. However, in order to leverage all the advantages of this class ofmaterials, polymer science must face a number of challenges and changes. The ris-ing cost of crude oil is currently changing economics in favor of polymers basedon renewable and sustainable feedstocks (biopolymers for injection molding costabout d1.60/kg) and it is reasonable to anticipate that the development of newmaterials from these sources will accelerate in the future and the rapid discoveryof new polymeric entities will become a strategic priority [27]. Furthermore, onemight speculate, that an increasing amount of work will be done on non-commodity,value-added applications of polymers, which are usually developed at the interfacesof several scientific disciplines: polymers and medicine, polymers and electronicsand polymers and nano-engineering.

Another significant driver of change is the increasing use of combinatorial andhigh-throughput experimentation (HTE) in virtually all areas of polymer science,starting with synthesis and discovery [28–34], processing [35–40] as well as screen-ing for chemical, physical and mechanical [41, 42] properties. While HTE has beenan integral and indispensable part of the pharmaceutical industry for a significantperiod of time, the materials science community has mainly persisted in using “one-step-at-a-time” experimental approaches. Recently, however, a significant effort toutilize combinatorial and high-throughput technologies has been made by a numberof industrial and academic laboratories, with the result that HTE is now enteringstandard laboratory practice.

One final force, which is currently revolutionizing science – and thereforepolymer science – is the world wide web (WWW) itself. The advent of the semantic

110 N. Adams

web is currently profoundly changing the way in which scientists communicateand exchange data and it can be expected that the impact of this will also touchpolymer science.

The discipline of informatics can loosely be defined as “the science of informa-tion” and is invariably coupled with both the engineering of information systems andthe practice of information processing. For our purposes, this defines the scope ofwhat polymer informatics should be and therefore also the scope of this contribution.

It has already been mentioned that one of the drivers of change in polymer sci-ence is an increasing amount of interdisciplinarity. From an informatics point ofview, “interdisciplinarity” most often denotes an exercise in data integration. Whenattempting to develop a polymer pharmaceutical, for example, not only does thepolymer chemistry have to be right, but during the development cycle, data fromthe synthesis laboratory must be integrated with data from other areas such as (clin-ical) pharmacology, toxicology, cellular biology, (pharmaco)genomics, etc. Whilethe problem is easy to state, cross-disciplinary data integration is far from trivial. Togive an example, does the term “macromolecule” mean the same thing when usedin the context of (polymer) chemistry and in the context of biology (in the formercase “macromolecule” often refers to a synthetic macromolecule which is a mem-ber of an ensemble of molecules which, together, make up a “polymer”, whereas inthe latter case “macromolecule” is often synonymous with “protein”)? And there-fore, what about data that is annotated with this single concept, which may havedomain-specific semantics? Similarly, is the definition of “glass transition temper-ature” in the context of polymer science the same as in the context of mineralogy?How can we get a computer to make that decision, so that it can decide whetherdata from a “polymer context” and data from a “mineralogy context” is equivalentand can be combined? Interdisciplinary (polymer) science forces us to address allof these questions and hence part of the remit of polymer informatics must be todevelop technologies, information architectures and ontologies that enable and fa-cilitate data exchange between several different disciplines.

The increasing adoption of high-throughput and combinatorial experimentationapproaches, too, triggers the need for sophisticated informatics support. While“high-throughput” in materials science usually implies experimental volumes,which are much smaller than those encountered in medicinal chemistry or drug dis-covery (HTE in polymer science typically means tens or hundreds of samples, ratherthan thousands or tens of thousands), materials HTE nevertheless generates signifi-cant amounts of data, which need to be administered, handled and stored. Once thenecessary metadata requirements such as measurement conditions, etc. have beencompounded into this, data volumes increase even further. The implication is, thatthe ways in which (academic) polymer science handles (by laboratory notebookand digital spreadsheet) and disseminates (by journal publication) polymer data donot scale or are inappropriate. Moreover, polymers are materials which are signifi-cantly more complex and fuzzy than small molecules. Hence, informatics solutions,which have been developed for the latter – both in the context of high-throughputand “classical” experimentation – do not translate well to the former: polymersrequire their own solutions. Finally, while the outcome of a HTE campaign is raw


experimental data, data on its own is quite useless. It is knowledge – the confidenttheoretical and practical understanding of a domain – which enables a scientist tomake decisions or to develop a product. Polymer informatics, therefore, needs todevelop technologies, which (1) allow the description of polymeric materials inan appropriate manner, (2) allow the attachment of knowledge and data to thesedescriptions and (3) allow the easy conversion of such data in knowledge.

It is worth noting that (scientific) data has intrinsic value, even if the primarypurpose for which it was created, has long been fulfilled. Innovative (re-)uses ofpublicly available data on the internet exemplify this nicely. Scientists usually tendto produce data in the context of a specialized research project and disseminateit through the means of scientific publication in a learned journal. Once researchobjectives have been met or a publication has been published, scientists often loseinterest in their own data as it serves no primary purpose anymore. Because of this,significant amounts of valuable scientific data either never get published and thusnever become part of the “knowledge commons” or are rendered inaccessible tomachines and thus effectively destroyed for informatics purposes.

By contrast, the modern internet demonstrates clearly how data that have beenpublished openly on the web in a machine-friendly form and subsequently “mashed-up”, i.e., been brought together with data from other sources or domains, cangenerate valuable new knowledge. Scientists are now beginning to tap into this vastpotential and to generate new science and insights [43, 44]. Polymer informaticsshould develop technologies to enable (polymer) scientists to publish and store theirdata in a way that is free of access-barriers (which can be technological, financial(e.g., subscription fees) or legal (e.g., attempts to copyright data)) and that allowsthe easy generation of mash-ups with data from other sources.

While considerable progress has been made in the area of small molecule infor-matics over the past several decades, any effort in the field of polymers has beentimid at best and there is considerable scope for development. The main reason forthe virtual non-existence of polymer informatics is the complex nature of polymers.This review will therefore start with an examination of the particular informaticschallenges posed by polymers, in particular in the area of polymer representationand will also discuss some of the peculiarities of polymer information (“the scienceof information”). It will look at information systems for polymers (“engineering ofinformation systems”) and a final section will review attempts to develop structure-property relationships for polymers (“practice of information processing”). Themodeling of polymers either on the molecular – or meso-level – is outside the scopeof this review.

2 Polymer Information is Challenging

The central dogma of chemistry is, that the structure of a molecule correlateswith its physico-chemical properties. This is usually illustrated using the corre-lation between the boiling point of n-alkanes and the number of carbon atoms

112 N. Adams

Fig. 1 The correlation between the boiling points of n-alkanes and number of carbon atoms in themolecule

in the molecule (Fig. 1). Knowledge of the structure of the molecule allows thecalculation of “descriptors”, i.e., variables, which capture information about aspectsof the structure of the molecule and which can be correlated to an observed physico-chemical property.

Historically, most chemists have modeled the structure of molecules using ahighly idealized “platonic” representation, where atoms are represented as ver-tices and bonds as paths between vertices. Chemoinformatics has very successfullyadopted this representation and based many of its techniques around the metaphorof the “connection table”, i.e., a list of all atoms and bonds, which occur in themolecule. While this approach is quite successful for well defined chemical enti-ties, it begins to break down for rapidly interconverting isomers, for example, andis completely inappropriate for polymers. In the majority of cases, the successfulapplication of chemoinformatics to a given problem depends on the availability of aconnection table.

For synthetic polymers, the representation of structure by connection table es-sentially breaks down because polymers are collections of macromolecules, eachof which is slightly different, even if only in terms of the degree of polymeriza-tion. If the central dogma of chemistry holds, though, this means that any physicalquantity observed as a “polymer property” is an average over an ensemble of macro-molecules. A polymer can therefore not be described by a single connection tablealone. Structural heterogeneity in synthetic polymers is virtually a given: even themost well-controlled polymerization reactions yield polymers with polydispersityindices (PI) of around 1.03 (see, for example, [45]). Because no single connec-tion table is capable of representing a polymer and because the observed physicalproperties of a polymer sample are the metrological sums of the properties of the


contributing entities, the connection table/data correlation approach breaks down inthe case of polymers and many other materials for that matter.

Furthermore, we often know how a polymer was made, but do not know thestructure of the resulting product – a famous example of this is Bakelite (phe-nol/formaldehyde resin). Conversely, as is often the case for industrial polymers,we may have some property data, but do not know how the polymer was made orany structural details. Polymeric materials in general have a “history” and, more of-ten than not, it is that history which influences or completely determines its propertycharacteristics and not the structural characteristics. The implication of this is, that adescription/encoding of a polymer’s history is often more relevant than an encodingof its structure.

A final factor, which complicates polymer data is the fact that crucial meta-data are often missing. A significant number of polymer properties are not onlydependent on the chemical nature of the constituent macromolecules of a poly-meric ensemble, but also on factors such as measurement methods, measurementconditions, etc. The glass transition temperature (Tg), for example, is formally con-ditional on quantities such as pressure, molecular mass, tacticity and cross-linking.For low molecular weight polymers, Tg increases with increasing polymer molec-ular weight until an upper limit is reached and the value of Tg becomes essentiallyinvariant to further increases in molecular weight. Furthermore the experimentalmethod used to determine the glass transition temperature is of crucial importance:popular measurement methods include differential scanning calorimetry (DSC) orthermo mechanical analysis (TMA). DSC determines the change in heat capacityof the sample as a function of temperature, whereas TMA measures dimensionalchanges (length and thickness) of a sample. If, therefore, one were to determine theglass transition temperature of a polymer sample using the two different techniques,the results obtained would usually differ by several degrees Kelvin.

The first significant challenge that polymer informatics has to tackle, there-fore, is to develop representations for both polymers and polymer data, which arecomputable, i.e., “machine-comprehensible” and to which rich metadata can beattached.

2.1 The Representation of Polymers

Early efforts in what could be termed “polymer informatics” go back to an ACSsymposium on polymer nomenclature in the late 1960s [46–51]. Papers in thissymposium were mainly concerned with issues of polymer nomenclature and as-pects of information retrieval. A first set of seminal papers only appeared abouta decade later, as a consequence of another ACS symposium on the retrieval ofpolymer information in 1978 [52–59]. Collectively, the papers resulting from the1978 symposium set out the challenges still faced by polymer informatics today:the fuzzy nature of polymers and the variety of different types of descriptions andrepresentations of polymers arising as a consequence, the problem of information

114 N. Adams

compartmentalization and the difficulty associated with information retrieval acrossthe numerous interdisciplinary interfaces of polymer science [52]. A contributionby Fugmann, describing the POLIDCASYR polymer documentation system of the“Internationale Dokumentationsgesellschaft fuer Chemie (IDC)” (a now defunctGerman chemical information service founded as a joint venture between BASF,Bayer, Hoechst, Degussa and Huels) is particularly relevant, in that it examinesboth the important syntactical and logical concept relationships between structuresand polymers and points to the use of controlled vocabularies and semantics forcapturing non-structural aspects. It furthermore introduces the notion of indexingpolymers as molecular fragments. The POLIDCASYR system is an extension ofthe GREMAS (Genealogical Retrieval by Magnetic Tape Storage) technology, alsodeveloped by IDC, which indexes molecules in terms of “molecular regions” (e.g.,molecular fragments). The GREMAS system distinguishes between four distinctmolecular regions, namely aliphatic carbon chains, alicycles, aromatic rings andheterocycles and uses term symbols to code for these. Typical terms include, forexample, symbols for large chains of carbon atoms of statistical length (term sym-bol “6”, Fig. 2), terms for carbon atom chains in non-polymers (term symbol:“YR”,Fig. 2) terms for hydroxy-, carboxy-, carboxylic acid etc. and many other groups.This leads to an indexing of molecules as a collection of molecular terms as shownin Fig. 2.

Another POLIDCASYR extension ensures that structural features of thebackbone can be distinguished from those occurring in side chains. The systemfurthermore augments structure codes for polymers using a controlled vocabularyof keywords, such as “epoxy resin”, “aminoplast” or “phenoplast”.

The POLIDCASYR paper also acknowledges the importance of the “history” ofa polymer and the oftentimes incomplete information associated with a polymericentity. This is important, because if the structure of a polymer is unknown, then anencoding of the material in fragment terms will be impossible and the information

Fig. 2 Examples ofGREMACS/POLIDCASYRterms for small molecules andpolymers

HOOH

HOO

OH

HO COOH

OH

CO2R

n

ROOC COCOR

n

"YRRE"

"YRRE.02"

"YRREN"

"Y6N"

"Y6EN"


A B A Bg

C

A B C g

BA

a b

e

c d

"A and B" "A and B in the process gyield C"

"A or B" "C in the process gyields A and B"

Ethylene Polymer Precursor

polymerisation

Polymer

Fig. 3 Basic elements of a TOSAR graph: (a) “A and B”, (b) “A and B in the process γ yield C”,(c) “A or B”, (d) “C in the process γ yields A and B”

scientist has little choice but to codify and index the polymer in terms of relatedconcepts, such as the monomer a polymer was derived from. This, in turn, requiresthe definition of a series of syntactical and logical relationships between conceptsfor successful information retrieval. To this end, the IDC developed the TOSAR sys-tem (Topological representation of Synthetic and Analytical Relations of concepts)[60], which represents the relationships between concepts by means of a graph andcan therefore be viewed as a precursor to modern semantic web and ontological ap-proaches such as, for example, the Resource Description Framework (RDF) [61, 62]or the Web Ontology Language (OWL) [63]. In a TOSAR graph, concepts are rep-resented as nodes and concept relations as edges of the graph. Concept relationscan exist as either “analytical relations” or “synthetic relations” (Fig. 3) and thus theTOSAR approach is, in essence, an incarnation of Ranganathan’s Colon Classifi-cation (CC) [64]. The colon classification is an ‘analytico-synthetic classification”,or, in modern parlance, a “faceted” classification. Furthermore, the TOSAR systemallows the specification of logical operations such as “AND”, “OR” and “ONLY”as well as optionality, which allows an indexer to effectively encode the history ofa polymer as well as the processes (e.g., synthetic reactions, grafting, sequences intime, etc.) described in a document (Fig. 4). In a sense, the resulting graph is botha structural representation of the information source it was taken from as well as arepresentation of the history of a polymer.

In addition to fragment and graph indexing of polymer information, the POLID-CASYR system also makes use of two distinct vocabularies for non-structural terms.The first vocabulary is, in essence, a controlled vocabulary of hierarchically or-dered terms (taxonomy), supplemented by a second, more fluid vocabulary, whichis subject to constant editing. The latter is used to further enhance the controlledvocabulary, e.g., the term “isomerization”, which is part of the controlled vocab-ulary, could be defined further by the terms “racemization”, “tautomerization” or“rotation isomerization”. Annotation of this kind is only a short step away fromtechniques, which we now associate with the terms “tagging” and “folksonomies”and which are typical components of Web 2.0 systems. POLIDCASYR’s con-trolled vocabulary is structured according to a number of semantic categories such

116 N. Adams

Acrylonitrile polymer precursor

Acrylic acid polymer precursor

Vinylacetate polymer precursor

(Co)polymerisation

Polymer F(acrylic resin)

Dissolving

Solution

Dimethyl-formamide(solvent)

H solvent G solvent

Acrylic Fbre

Dry-spinning

Fig. 4 TOSAR graph representing a sequence of processes

Properties

Process-related non-process-related

inhibitors

inhibitors

influencers

catalystspassive

participants

active

apparatus

Fig. 5 Subdivision of semantic categories in the POLIDCASYR controlled vocabulary

as substances, apparatus, living entities, processes, reactions, process-related andnon-process-related properties and fields of application (Fig. 5).

These top level categories can be subdivided further and establish relations be-tween concepts (Fig. 5). Each of the subdivisions has a unique code, which can beused for indexing and searching.

Other papers from the 1978 symposium discuss the indexing of polymers inChemical Abstracts [53], the IFI/Plenum System [59] and an in-house system forHercules [58]. While a number of services such as the Chemical Abstracts Service


have several routes into polymer information (keyword search, search by name,search by formula index, search by general subject index), for most chemists, how-ever, the most important route to polymer information is either via “search by name”or “search by structural formula”.

2.1.1 Name-Based Representation of Polymers

Name-based representations of polymers normally occur in three different incarna-tions: (a) trade and trivial names, (b) names derived from the component monomersof a polymer (“source-based representations”) and (c) names derived from thestructural repeating unit of a polymer (“structure-based representations”). Triv-ial or trade names are often derived from the inventor of a certain material (e.g.,Bakelite – developed by Baekeland) and contain no or only little chemical infor-mation and therefore very easily lead to information compartmentalization – trivialnames usually cannot even be inferred. Which of the other naming conventions isused normally depends on the nomenclature philosophy of a particular communitywithin (polymer) chemistry and there is no general agreement as to which typeof nomenclature is preferable. Polybutadiene, for example, would be indexed as“1,3-butadiene, homopolymer” by the Chemical Abstracts Service (CAS) (Fig. 6).A keen observer will also have noticed that the name has been inverted, which isCAS specific. The International Union of Pure and Applied Chemistry (IUPAC)will allow the registration of the polymer as either “polybutadiene” (IUPAC source-based), “poly(but-1-ene-1,4-diyl)” (IUPAC structure-based), “1,4-polybutadiene”(IUPAC semisystematic), “poly(buta-1,3-diene)” (IUPAC source-based). Apartfrom particular idiosyncrasies such as name inversion or the use of brackets, nomen-clature systems are also subject to historical (dis)continuities: as time passes andindexing systems evolve, nomenclature rules change: poly(ethylene terephthalate)(PET) is registered as “poly(oxyethyleneoxyterephthaloyl) in the 8th CollectiveIndex of Chemical Abstracts, but as poly(oxy-1,2-ethanediyloxycarbonyl-1,4-phenylenecarbonyl) in the 9th Collective Index.

While different nomenclature conventions and historical discontinuities are atbest confusing to the human searcher, they lead to serious problems in terms ofmultiple registration of entities in registration systems as well as information com-partmentalization: a mapping of “polybutadiene” onto “poly(but-1-ene-1,4-diyl)”is not straightforward for a human and even less so for a machine (polybutadienedoes not indicate the position of the double bond in the polymer backbone) andtherefore any data associated with either registration may or may not be equivalent.

Fig. 6 Polybutadienerepeating unit

CC

H2

CCH2

H

H

n

118 N. Adams

Ontologies could go some way towards solving this problem since they assist in thespecification of parthood relationships (e.g., between a single concept and multiplelabels).

2.1.2 Structure-Based Representation of Polymers

Apart from simple name-based representations, polymer information systems alsouse chemical structure diagrams to register polymeric entities. We have alreadydiscussed the breakdown of the structure diagram metaphor for polymers in thesections above and the situation currently encountered in many polymer informa-tion systems is reflective of this: polymers are either registered using the structureof the corresponding monomer(s) or a greatly reduced representation in the form ofthe repeating unit of one of the macromolecular constituents.

Structure representations of polymers based on monomer structures are the leastdesirable, because they reference one concept in terms of another and necessarilylose all polymer-specific information: often, monomers undergo significant struc-tural change upon polymerization (e.g., going from cyclic to linear structures duringring-opening metathesis polymerizations (ROMP)) and potentially important in-formation such as endgroups, post-modification procedures such as grafting, etc.,cannot be encoded. Representation by repeating unit structure is problematic too:while the repeating unit is at least indicative of the major structural features ofthe polymer backbone (end group information, post-processing, etc., is still not en-coded), often the problem of unambiguous representation arises. When consideringthe structure of a poly-1,3-butadiene macromolecule, for example, it becomes evi-dent that several valid repeating unit structures can be drawn (Fig. 7).

The fact that several representations are possible automatically necessitates thedevelopment of rules which would allow a researcher to decide upon a preferred rep-resentation. To this end, IUPAC has developed an elaborate rules-based system usingthe “seniority of subunits”, the direction of citation, etc. [65]. However, rules-basedsystems are subject to the same limitations as nomenclature systems in that they,too, suffer from (potential) historical discontinuities and require acceptance by abroad community.

H

H

H

H

H

H

H

H

H

H

H

Hn n

H

H n

Poly-1,3-butadiene

a b c

Fig. 7 Possible repeating unit representations (a, b, c) for poly-1,3-butadiene


One paper arising from the 1978 symposium points out another central and un-solved problem in polymer informatics today, namely the scarcity of polymer data[56]. The paper suggests that three things are needed to develop a successful poly-mer information system: (1) the availability of good data models, (2) the long-termcommitment of both industry and academia to generate comprehensive and meta-data rich materials data, which adheres to the data model and (3) “the commitmentof national data centers to act as repositories of such information.” The importantmessage here is that the successful development of information systems for materi-als is mainly a “political” problem: communities of practice need to agree to produceand share data and to preserve and curate it for the long term. Good technology isessential, but is hardly ever the real limiting factor. For polymers, none of theserequirements have been met so far. However, developments on the WWW such asan increasing emphasis on community-developed or community-created informa-tion and data (e.g., Wikipedia and Freebase) together with developing technologiesfor linking and “mashing-up” data might go some way towards meeting the aboverequirements.

While the 1978 symposium on polymers highlighted many of the problems asso-ciated with polymer information and also discussed some possible solutions, paperssubmitted for another ACS symposium on the topic entitled “Polymer Information:Storage for Retrieval, or Hide and Seek?” suggest, that nothing much has changed[66].

One notable exception to this is a paper by Gushurst et al. introducing MDL’s(now Symyx) proprietary Sgroup notation [67]. “Sgroup” is an acronym for “sub-structure group” and represents a “persistent collection of atoms and bonds” whichis a part of a larger connection table. Sgroups subdivide into two types, “chem-ical Sgroups” and “data Sgroups”. The Sgroup notation can be used for the de-scription of polymers, formulations and other materials. Sgroup representations ofboth a structure-based (Fig. 8a) and a source-based (Fig. 8b) block copolymer ofpolystyrene and polyethylene oxide are shown in Fig. 8.

In Sgroups, repeating units are enclosed in square brackets and a subscript “n”is placed to the right of the closing bracket. The subscript “blk” indicates a blockcopolymer and “mon” a monomer in a source-based representation. Superscriptsindicate the orientation of the repeating units (hh = head-to-head, ht = head-to-tail, eu = either unknown) in a structure-based description. Crossing bonds (bonds

*O

n *n

blkmon

blk

O

mon

a b

Fig. 8 (a) Structure-based and (b) source-based Sgroup representation of a polystyrene/polyethylene oxide block-copolymer

120 N. Adams

crossing the square brackets) indicate known connectivity. Data Sgroups may beassociated with these structural descriptions and can hold information about, for ex-ample, the percent composition and distribution of building blocks in a polymer,modifications, etc. Sgroups therefore go some way towards being able to describethe history of a polymer as much as its composition. When searching for a particularpolymer using a structure-based representation, MDL’s proprietary “flexmatch” al-gorithm cyclizes all possible repeating unit representations. If multiple valid repeatunits can be written, the cyclization process creates a set of molecules, which are“phase-shifted” with respect to each other. This means that all structures are effec-tively identical, which removes the need for rules for choosing a preferred repeatingunit representation.

A paper by Kaback re-enforces the previously made point that “there’s more toa polymer than just its build” [68] and that it is the history of a polymer that isoften more important for information retrieval than any complex structure represen-tation. As an example, Kaback points out, that, for a polymer, it makes a significantdifference whether the polymer was synthesized in a stirred autoclave or a tubu-lar flow-through system, as this will lead to polymers of vastly different physicalcharacteristics, even if all else is equal. This information, however, is hardly everencoded as metadata in information systems. The same is true for other types ofmetadata – it is currently very difficult to distinguish information about elastomericethylene/propylene copolymers from data referring to, for example, rigid thermo-plastic ethylene/propylene copolymer, etc. In Kaback’s own words: “Polymers areboth chemicals and materials. In viewing a polyolefin as a chemical reactant, it iscritical for me to know if it was produced by a catalyst system that gives a productin which residual chain unsaturation is overwhelmingly at the chain end, rather thanone or more units into the chain. In viewing it as a thermoplastic, it is critical forme to know whether the molecular weight distribution is broad, implying good flowproperties, or narrow, suggesting possible melt processing difficulties” [68]. In otherwords, what determines the value of a polymer information system is the “axioma-tization” of the polymeric entity and the context-aware “lens” onto the informationthat the system should provide. Other papers in the 1991 symposium series are es-sentially descriptions of a number of proprietary information systems, such as theDerwent Plasdoc System [69], the IFI/Plenum Polymer Indexing System [70] theRapra Abstracts Rubber and Plastics Database [71], with side-by-side comparisonsof all these systems [72, 73] and a discussion of the particular challenges associatedwith searching for information on condensation polymers [74].

2.1.3 The Semantic Web of Polymer Data

Many of the challenges described so far can be addressed using technologies de-veloped in the context of the “semantic web” or “web 3.0” [75]. The semantic webis currently revolutionizing the way in which we structure, handle, present, and ex-change scientific data and information. Unlike the current incarnation of the web,which is mainly a web of documents, the semantic web is a vision of a web of


data, in which software robots are able not only to discover information, but alsoto understand its meaning and therefore to act on it autonomously. Currently, theway in which we communicate polymer information is similar to the way in whichwe treat information on the web. Most polymer information is contained in unstruc-tured documents (equivalent on the web: webpages) such as scientific papers, theses,patents. etc. Because these documents are unstructured (a machine does not a prioriunderstand that the symbol combination “◦C” denotes a temperature unless this isexplicitly stated in machine comprehensible language), it is extremely hard for ma-chines to discover information in these documents. Moreover, information about agiven polymer is often scattered across several sources (web analogy: many web-pages on different servers/sites talking about the same concept) and hard to combineinto a single picture. The semantic web is developing some technical solutions to ad-dress these problems in the web domain and it is reasonable to assume that these canalso be applied to the domain of polymer information.

In a typical scenario, a polymer scientist wishing to design a new polymer for agiven application would specify an application/property profile of that polymer. Asemantic web agent (software that acts on behalf of a user), in turn, would be able to“understand” the specification and to collect relevant information from the “cloud”(e.g., the web, in-house data sources, proprietary and open databases, scientific pub-lications, etc.). The agent might then reconcile the gathered information against therequirements profile and use existing quantitative structure-property relationshipsor rules to infer other properties. Finally, the agent would present a list of polymerswith a property profile closely matching the original specification. To achieve thisvision, a software agent would not only have to find information (i.e., informationmust be present in a structured form), but also understand its meaning (i.e., the in-formation must be computable). On a technology level, therefore, we require thethree components of (1) publication (data needs to be available digitally), (2) se-mantics (data needs to be semantically rich), and (3) data interoperability (whichcan be achieved via the use of semantics).

The technological foundations of this vision currently consist of eXtensibleMarkup Language (XML) [76], XML Schema [77], the Resource DescriptionFramework (RDF) [61], RDF Schema [62], and the Web Ontology Language [63].These technologies are interdependent and can thus be arranged in the form of a“semantic layer cake” (Fig. 9).

One way to impart structure to otherwise unstructured documents is to utilize asuitable markup language. The function of markup languages is to combine the textof a document with further information about the text (markup languages typicallyadd “metadata” – data about data) and while metadata is normally hidden fromthe view of a human reader, it is available to processing software. XML allows anauthor to add arbitrary metadata to documents through the use of tags, which areuser-defined and annotate data sources.

With XML providing the mechanism for structuring, the Resource DescriptionFramework (RDF) provides the machinery for data integration and lightweightaxiomatization. A “resource” in informatics terminology is anything that can benamed, addressed, or handled and the language provides a framework for making

122 N. Adams

Fig. 9 The semantic layer-cake. (Copyright 2008 World Wide Web Consortium, (MassachusettsInstitute of Technology, European Research Consortium for Informatics and Mathematics,Keio University). All Rights Reserved. http://www.w3.org/Consortium/Legal/2002/copyright-documents-20021231. Reproduced with permission)

statements about these resources. Statements are made in the form of “triples”,i.e., almost human-language like subject-predicate-object constructs. RDF thereforeallows simple assertions of the type: “polystyrene is a polymer” to be made. Theweb ontology language (OWL), finally, extends this expressivity even further byallowing the addition of first-order-logic. This, in turn, means that relationships be-tween resources, such as disjoints, cardinality, equality and symmetry, etc., can bedescribed.

Semantic technologies are highly significant for polymer (data) representation,because they are now being applied in the field of chemistry and polymer science.As XML is indeed extensible, a number of markup languages have been devel-oped over the last decade, which are of relevance in the general area of chemistryand which can also be used to structure polymer information. The most signifi-cant of these languages are Chemical Markup Language (CML) [78–83], AnalyticalMarkup Language (AnIML), and ThermoML (markup language for thermochem-ical and thermophysical data) [84]. Some papers detailing the beginnings of analternative to CML as a general chemical markup language have also recently beenpublished [85, 86].

Chemical Markup Language was pioneered by Murray-Rust and Rzepa and isdesigned to manage mainly molecular information, such as chemical structure aswell as spectral, analytical-, computational-, and crystallographic data. CML iscurrently being investigated by Microsoft Research as a way of introducing semanti-cally rich chemistry features within the Microsoft Office Word package [87]. Adamsand Murray-Rust recently reported the development of Polymer Markup Languageas a polymer-specific extension to CML [88, 89]. Like the Sgroup approach, the


language is a fragment-based representation and addresses considerations such asthe composition of a given polymer, the structure of a polymer or macromolecule(if known), as well as the records of a computational experiment, physical or otherproperties of a polymer or macromolecule, experimental metadata, and reactioninformation. At the moment, PML does not consider polymer processing such ascompounding, but may well be extended to cover these areas in the future. The de-sign criteria of PML include the re-use of CML whenever possible, full namespaceawareness, interoperability with other mature STM languages, the avoidance of im-plicit semantics, the treatment of polymer-specific phenomena such as the ensemblenature of polymers, ambiguous repeating units, tacticity, double bond isomerism,macromonomers as well as the description of all commonly encountered struc-tural motives such as homopolymers, copolymers (statistical, alternating, block),branched polymers (combs, hyperbranched systems, dendrimers) as well as cross-linked polymers and graft polymers. Figure 10 shows a PML document for a simplestyrene heptamer.

The construction of the molecule starts with a root fragment which, in the partic-ular document in Fig. 10, is a methyl group (<molecule ref = "g:me">).The “g:me” pointer links to a fully atomistic description of the methyl fragment inCML, which in turn allows the annotation of individual atoms if needed – this couldbe important when wishing to specify differential reactivity of backbone atoms forgrafting and post-processing, for example. The methyl group joins onto a CH-group(<molecule ref = "g:ch">), which is connected to a methylene fragment(<molecule ref = "g:ch2">) and a phenyl fragment (<molecule ref= "g:benzene">). The construction thus defines a full repeating unit, thoughthis is completely accidental and not required by the language – the user is freeto choose any type of fragment, such as multiple repeating units, macroinitiatorfragments, etc.

The <fragment> container, which is a child of the root element, con-tains a countExpression attribute, which indicates to PML processingsoftware that the fragment definition is to be repeated another six times. ThecountExpression attribute is, in fact, a computable variable. This is a newconcept for markup languages, which have hitherto been used in a completelydeclarative way. The countExpression element can evaluate either determin-istically (specifying a macromolecule) or stochastically (specifying distributions ofmacromolecules). In general, PML can either be implicitly or explicitly computed,which means that it is possible to create documents containing free variables. Assuch, it is possible to embed XSL Transformations [90] into PML documents andto evaluate the resulting expressions in a lazy manner.

PML can also represent polymers at different levels of certainty: where the struc-ture is known, it can be encoded and where this information is not available, apolymer can be codified in terms of other concepts, such as the monomers it wasprepared from. Furthermore, PML provides essentially a coarse-grained representa-tion: larger structural fragments can be mapped back to fully atomistic fragments,if desired. PML, therefore, allows data to be associated with a polymer representa-tion at the atom, molecular fragment, molecule and molecule ensemble level, and,

124 N. Adams

Fig. 10 A simple PML document showing the construction of a styrene oligomer

as such, it can be considered to be a “normal” (though not a canonical) form ofa polymer description. The authors have recently published a “polymer builder”prototype application, making use of polymer markup language and the associatedprocessing software [88]. As part of one of their papers, Adams et al. also brieflypresent the development of polymer ontologies, which contain basic polymer sci-ence terms, although no formal publication has appeared yet. To date, ontologiesare severely underused in chemistry and chemical data management, although thereare a few attempts at constructing ontologies in the domain. Probably the mostprominent chemical ontology is the European Bioinformatics Institute’s “Chemical


Entities of Biological Interest” (ChEBI) [91]. ChEBI is a small-molecule “database”of natural products or synthetic entities, which have been used to interfere in biologi-cal processes and encompasses an ontological classification specifying relationshipsbetween molecular entities. Information about chemical entities is derived from anumber of datasources, such as COMPOUND [92], the Chemical Ontology (CO),and IntEnz [93]. Other EBI ontologies in the chemistry domain are REX [94]and FIX [95], which describe physicochemical processes and methods respectively.Other groups have also reported efforts to model chemical structure [96], reactions[85, 86], and laboratory processes [97–99]. Further ontologies for the general sci-entific domain encompass ontologies of the scientific experiment as such [100], aswell as a number of upper ontologies, which are suitable for science (e.g., SUMO[101], General Formal Ontology [102]). When taken together, it is reasonable toassume that these will be useful for describing and capturing much of the metadatathat many of the “early” polymer informaticians requested from their informationsystems and that concepts from these ontologies can also contribute to the develop-ment of computable (i.e., “machine comprehensible”) definitions of what a polymer,a glass transition temperature, etc., is.

A typical example of the possible definition of a macromolecule in the OWLontology language is shown in Fig. 11. Here, a macromolecule is modeled in termsof sets: a macromolecule is a molecule which has a high relative molecular mass andwhich is derived from either a molecule of low molecular mass or from an oligomeror from molecules of low molecular mass and oligomers. It is now easy to see howone can develop axiomatizations such as those envisaged by the TOSAR system(see above) using modern semantic web technologies and how information aboutthe history of a material can be encoded in a machine comprehensible way, whichis subsequently available for reasoning and knowledge discovery [88].

2.2 Access to Polymer Information

So far, all of the discussion in this review has focused on the representation of poly-mer structure and polymer information. However, another significant challenge inthe development of polymer informatics is access to polymer data. In this context,the term “access” takes on two distinct meanings, namely “access” to data in termsof access-barriers (e.g., proprietary data, copyright considerations, etc.) and “ac-cess” in terms of the formats in which polymer data is communicated, handledand exchanged.

2.2.1 Access and Technical Access-Barriers

The first and obvious technical access barrier is the availability of data in digi-tal form. While most modern documents are “born digital”, a lot of libraries willstill only archive paper copies of scientific literature, although more and more

126 N. Adams

Fig. 11 Possible ontological description of the concept “Macromolecule” in the OWL ontologylanguage

institutions now require documents to be submitted to institutional repositories indigital form. Furthermore, the increasing use of open access and self-archiving man-dates by universities contributes to the availability of data. However, this is a gradualprocess and it is only now that a growing number of university libraries engage inthe development of institutional repositories. Some sources of polymer information,such as the Polymer Handbook [103], are available on paper only.

Even if a document containing polymer information is available digitally, the for-mat of the document is critical. While most scientific documents are authored usinga word processing system such as Microsoft Word or Open Office or a typesettingsystem such as LaTeX, this is not normally the format they are subsequently dis-seminated in. In the majority of cases, these documents are converted to portabledocument format (pdf) for publication and dissemination. It is this very step whichis problematic for automated information extraction. Portable document format hasbeen designed as a purely presentational format: “The PDF design is very tailored tothe creator being able – quite directly and without ambiguity – to specify the exact


output desired” and even its developers admit, that it is extremely difficult to extractinformation [104]. Essentially, pdf converts text to a set of graphical objects withoutsemantics, i.e., there are no well-defined relationships between these objects andthe conversion process is “lossy” in terms of information content. A consequenceof this is, that a lot of vital scientific information is lost when converting back totext: frequently subscripts and superscripts disappear, a loss of vital informationwhen attempting to teach a computer to recognize molecular formulae in text, anddata tables are often completely destroyed. An example of this is a recent paper byFeniri et al. in which the authors reported the synthesis and spectroscopic character-ization of a large library (630 compounds) of polystyrenes [105]. Each compoundin the library was characterized by both IR and Raman spectroscopy and the re-sulting data was published as spectral plots in pdf format. The result of this formof data publishing is that the spectral data associated with the polymers is lost: thedata is only useful for human inspection and effective machine learning and datamining as well as the possibility of generating mashups with other data sources hasbeen destroyed. In this particular case, the authors had little choice but to publish inthis format (although archival of the raw spectral data in an institutional repositorywould also be an option): the chemical and polymer science community has simplynot yet evolved mechanisms for publishing its data and not “just” its papers.

Even if the document is produced in a format suitable for information extractionand does not of itself destroy information, it is often the reporting scientist who does.A good example is the way in which NMR data of organic molecules are currentlyreported: the digital signal obtained from the NMR machine, even when processed,contains significantly more information than is reported using the standard journalpublication procedures (i.e., peak positions, splitting patterns and coupling con-stants). It is not possible to reconstruct the original spectrum from this data andany information that may have been contained in the original signal and which isnot captured in the reported data is lost. In the era of digital documents, virtuallyfree and unlimited storage and pervasive computing and network access, scientistsshould be much less willing to simply throw data (and thus potentially information)away, in particular if the data has been produced at great cost. (Chemical) Crystal-lography is exemplary in this regard: as a community, the discipline has evolved datastandards (cif file format [106] and the crystallographic information framework) andtools and mechanisms for data preservation, sharing, and archiving (e.g., the Cam-bridge Crystallographic Database [107], CrystalEye [108]). There are currently notechnical obstacles to doing the same for polymer data. What is severely lacking,though, is an understanding of data (“I am a chemist and shouldn’t have to worryabout data (formats)”) as such, a willingness to become educated, and communityagreement on how to deal with scientific data.

Another challenge to data access is presented by the unstructured nature of thedata contained in many documents. From “the point of view of a computer”, anunstructured document is a collection of symbols without any semantics, i.e., thephrase “polystyrene has a glass transition temperature of 95◦ C” is, in the absenceof further structuring data, i.e. metadata, meaningless and even the presence of datais hard to detect.

We have already discussed that a technical solution to this is to provide metadatain the form of markup. Markup can be introduced into a document either at the time

128 N. Adams

of authoring (a priori) or once the document has been prepared (a posteriori). Asthe authoring of markup is potentially complex, good semantic authoring tools arerequired. At the moment these are largely absent, although the Chem4Word projectaims to fill this gap [87]. If markup is introduced a posteriori, one usually has to en-gage in a significant amount of “information archaeology” and utilize technologiessuch as natural language processing, entity extraction, and parts-of-speech tagging.An example of this is the OSCAR 3 system, which is currently being developed byCorbett and Murray-Rust and extended to polymers by Jessop and Adams [109].OSCAR 3 is an open and extensible system for the automated markup of chemistryin documents. The markup is XML-based and designed to support browsing andchemistry-aware searching.

Table 1 shows an example of markup, generated using the OSCAR 3 system. Theabstract of a polymer research paper has been parsed by OSCAR and the resultingmarkup for the first sentence of the abstract is shown in-line with the text (Table 1B).The first chemical entity encountered in the sentence is “oleic acid”, which has beenmarked up as type = CM (Chemical Moiety) and a number of other annotations,such as in-line representations of chemical structure (InChI, SMILES) have beenattached.

2.2.2 Access and “Political/Cultural” Access-Barriers

While any technical problems associated with access to polymer data can be over-come with comparative ease, the real access problem is politico-cultural in nature.We have already alluded to the fact that scientists often produce data for the single

Table 1 An abstract ([174]) (A) prior to markup, (B) after markup with OSCAR 3

(A) Elaboration of PLLA-based superparamagnetic nanoparticles: Characterization, magnetic be-havior study and in vitro relaxivity evaluation Abstract. Oleic acid-coated magnetite has beenencapsulated in biocompatible magnetic nanoparticles (MNP) by a simple emulsion evaporationmethod.

(B)<?xml version = "1.0" encoding = "UTF-8"?>

<PAPER><TITLE>Elaboration of PLLA-based superparamagneticnanoparticles: Characterization, magnetic behaviourstudy and in vitro relaxivity evaluation.</TITLE> [175]<ne surface= "Oleic acid" type= "CM" provenance="unknown" SMILES = "CCCCCCCC\C=C/CCCCCCCC(O)=O"InChI="InChI=1/C18H34O2/c1-2-3-4-5-6-7-8-9-10-11-12-13-14-15-16-17-18(19) 20/h9-10H, 2-8, 11-17H2, 1H3, (H,19, 20)/b10-9-" cmlRef="cml1" ontIDs="CHEBI:16196">Oleicacid</ne>-coated<ne surface="magnetite" type= "CM"provenance="nGramScore" weight="0.09220993385201925">magnetite</ne>hasbeen encapsulated in biocompatible magnetic nanoparticles (MNP)by a simple emulsion<ne surface="evaporation" type= "ONT"provenance="oscarLexicon" ontIDs="REX:0000178">evaporation</ne>method....</ABSTRACT>


purpose of accomplishing the aims of the research project they are working on atthe time of data production. Once this has been achieved, the data produced duringthe project becomes virtually obsolete and its main value for the scientist lies in theproof of the fact that project work has been carried out at all and that it has beencarried out with scientific integrity [110]. There is very little return in taking anyfurther steps beyond “standard” journal publication to ensure that his data is easilyaccessible and reusable by other scientists, even when there are no commercial orother obstacles to data sharing.

Apart from very simple considerations of return or reward, the very system of sci-entific publishing, which science has evolved, now represents a de facto obstacle todata sharing and dissemination. Scientific publishing is historically a response to ascaling problem: in the early days of modern science, data and results were commu-nicated through letters sent between scientists. With the growth of the community ofscientists and the breadth of scientific endeavor, letter writing clearly did not scaleany more and the twin institutions of the “learned society” and the scientific journalpublication were developed [111]. The remit of the scientific journal was to col-lect manuscripts, to organize some form of verification of the “reasonableness” ofthe publication’s content and to print and distribute the resulting journal issue. Thejournal subscription fee formed the economic basis of the system. Much of the valueproposition of current STM publishing is still rooted in this business model. Withthe advent of the digital document, the internet and the set of technologies currentlysubsumed under the “Web 2.0” label, the economic foundation for this model hasall but collapsed and the internet now fulfils many of the functions a traditional pub-lisher used to perform. Document collection can be automated (users upload theirdigital manuscripts into an electronic workflow) as can, in principle, peer review(a weblog (“blog”) with a commenting function is nothing other than a publicationsystem with built-in peer review) and, of course, virtually cost-free publication anddistribution can be achieved by simple publication of papers on a website. Pervasivecomputing and networks mean that the economic cost associated with each of theseactivities is minimal and tending towards zero. Publishers are therefore increasinglyattempting to shift their value proposition to content and data: the subscription to anelectronic journal is often redefined as a subscription to a database. This redefini-tion is particularly problematic because once the subscription is discontinued, so isaccess to the content. Physical copies of paper-based journals, by contrast, remainin the subscribing institution’s library even after discontinuation of the subscriptionand access is therefore secure. The shift to content as the main value proposition isfurthermore exemplified by what appears to be an attempt to apply copyright state-ments to factual data, which is now effectively preventing scientists from accessingtheir own data in all but the most technologically backward way.

Copyright was originally conceived to protect the property rights of an author andto facilitate the governance of the use of an expression of an idea, but neither theidea itself nor factual data. Although copyright subsists in the particular expression

130 N. Adams

of the idea of an anthropomorphic mouse (e.g., Mickey Mouse), copyright law doesnot stop other parties from developing different expressions of the same idea andfrom subsequent exploitation.

This discussion is relevant here, because the act of publication of a scientific pa-per containing polymer data means that the publisher merely owns the expressionof the facts and data, but not the fact or the data itself. Unfortunately, publishersincreasingly appear to attempt to claim copyright on scientific data by attachingcopyright statements to, for example, supplementary data which is almost com-pletely factual. In an optimistic interpretation of these practices, one might assumethat it only gives an impression that the data is copyrighted and the copyright state-ment merely refers to the layout of the supplementary information. In the worstpossible interpretation, this can be viewed as a data grab by the publishers, who aretrying to erect barriers to data access which universities or private institutions wouldnot have the appetite to test in court. The practice of appending copyright statementsto supplementary information in any case obfuscates the data access situation. Otherdisciplines, such as biology, environmental science, and physics, have long evolvedboth a discipline and an ethics of data sharing and open access to data. Althoughthere are examples to the contrary (e.g., PubChem), chemistry and polymer scienceare astoundingly backward in their approach and attitude to caring about their dataas well as to sharing and dissemination. Ultimately, the successful development ofpolymer informatics will require access to polymer data and hence a culture of “dataawareness” and sharing needs to be developed in the polymer science community.

3 Making Use of Polymer Data

Another aspect of polymer informatics, beyond the representation and registrationof polymer information and data, is the conversion of data into knowledge and thusinto the power to make decisions. To this end, the same tools, which are common insmall molecule informatics, have also been used to study polymer data. The workthat has been reported so far subdivides into two categories, namely classificationand chemometrics problems and property prediction.

3.1 Classification and Chemometrics Problems

The rapid classification of polymeric species is an important problem in the area ofanalytical chemistry in general and of particular relevance to recycling and wastemanagement. To accomplish classification tasks, a combination of spectral data andprincipal component analysis (PCA) is often employed.

Principal component analysis is a simple vector space transform, allowingthe dimensionality of a data set to be reduced, while at the same time minimizing


the concurrent information loss [112–114]. In essence, PCA transforms data fromone coordinate system to another in such a way that the greatest data variance comesto lie on the first axis of the new coordinate system (the first principal componentwith the highest eigenvalue), the second largest variance on the second axis (thesecond principal component) and so forth. There are as many principal componentsin the system as there are dimensions. Data reduction is achieved by discardingcomponents with variances below a certain threshold.

Van den Eynde and Bertrand described the use of PCA for the determina-tion of molecular weights from time-of-flight secondary ion mass spectrometry(ToF-SIMS) spectra of polystyrene samples [115]. ToF-SIMS spectrometry is wellestablished for the characterization of polymer surfaces, but the quantification ofthe resulting spectra has proved to be difficult. In a previous study, the authors hadrecognized an influence of the molecular weight of their polymer samples on theintensities of observed secondary ions. In the present study, the researchers used18 polystyrene samples of different molecular weights and different butyl end-groups (n-butyl, sec-butyl, tert-butyl), which were spin-coated onto silicon wafersand subsequently subjected to ToF-SIMS analysis. After pre-processing, the spec-tral data was subjected to PCA. The results showed, that the first two principalcomponents were sensitive to both the molecular weight and the chemical struc-ture of the endgroups. The most significant ions in the spectrum can be detectedfrom the corresponding loading plots and the score plots allow samples to becategorized in terms of end-group structure and molecular weights. Furthermore,a universal calibration curve could be found by plotting the first principal com-ponent, which is independent of the structure of the endgroup, as a function ofmolecular weight. This can then be used for the determination of the molecularweight of an unknown polystyrene sample from its secondary ion mass spectrum.A similar approach was reported for the characterization of hyperbranched aliphaticpolyesters [116].

Batur et al. also used a number of machine learning tools, including PCA, anda set of different experimental techniques in a bid to quantify the crystallinity oflow-density polyethylene (LDPE) films [117]. In a multistep experiment, a first setof training data was produced by heating a thin film of LDPE to 120◦C in orderto produce a completely amorphous sample. The sample was then cooled in stepsof 2◦C and a Raman spectrum was recorded at each temperature step. The spec-tra were subsequently used as inputs for principal component analysis and neuralnetwork modeling. The inputs were correlated to crystallinity values derived fromsmall angle light scattering (SALS) experiments, which were carried out at the sametime and calibrated by DSC. The authors developed a linear regression model ofcrystallinity by correlating the factor loadings of the significant factors to the exper-imentally determined crystallinity values. It could be shown, that models determinedin this way validated well with respect to experimental data and also predicted thecrystallinities of test samples subjected to different cooling rates with satisfactoryaccuracy. Modeling using neural networks led to comparable results.

Miranda et al. used a combination of Fourier-Transform Infrared Spectroscopy(FT-IR) and PCA to elucidate the chemistry that occurs when poly(vinyl alcohol)

132 N. Adams

(PVA) is cross-linked under ultraviolet (UV) irradiation using sodium benzoateas an initiator [118]. An aqueous solution of PVA and sodium benzoate was castonto a glass plate and the solvent was allowed to evaporate. After irradiation ofthe samples for 1, 2, 3, and 4 h, the IR spectra of the films were determinedand analyzed by PCA. A detailed analysis of the scores and loadings led to theconclusion that upon irradiation the sodium benzoate is decomposed in an ini-tial step. The resulting radical subsequently abstracts a hydrogen atom from thePVA chain to produce a polymeric radical. The latter reacts with available hy-droxy groups to form ether bond linkages between PVA chains. Furthermore, itwas found that there is a good linear correlation between the mean of the scoresof one of the significant principal components and irradiation time, which meansthat the irradiation time of an unknown sample can be determined from the IRspectrum.

The use of PCA for the classification of both natural and synthetic polymerswas demonstrated by Vazquez et al. [119]. In their work, the researchers recordedTotal X-Ray Fluorescence (TXRF) spectra of scleroglucan, xanthan, glucomannan,poly(ethylene oxide), and polyacrylamide and subjected the resulting spectral datato PCA. To the naked eye, the X-ray fluorescence spectra of the polymers lookvirtually identical. However, when subjected to PCA it could be shown that thefirst two principal components contain approximately 96% of the variance in thedataset. When plotting the scores of the two components against each other, sixdistinct clusters are observed, which clearly differentiate the individual polymers.

Principal Component Regression (PCR) was used by Tuchbreiter and Muelhauptto determine the composition of a number of random ethane/propene, ethane/1-hexene, and ethane/1-octene copolymers [120]. After polymerization, the polymerswere characterized by both Attenuated Total Reflection Fourier Transform InfraredSpectroscopy (ATR-FT-IR) and 13C NMR and multivariate calibration models usingPCR were subsequently developed to estimate the co-monomer content.

The data generated by other experimental techniques is also amenable to decom-position/analysis by PCA. Lukasiak et al. reported the use of several classificationtechniques, such as PCA, k-means and hierarchical clustering, linear discrimi-nant analysis, k-nearest neighbors as well as distance metrics using Euclidean andMahalanobis measures on data generated by Dynamic Mechanical Analysis (DMA)of several types of polymers (polypropylene, LDPE, polystyrene, acrylonitrile-butadiene-styrene) of several different grades [121]. In their paper, the authorsdetermined the damping factor (tan δ) of the polymers as a function of tem-perature and showed that this data, in combination with several of the machinelearning techniques listed above, can be used to classify polymers into types andgrades.

All of these studies suffer from the fact that they were carried out on relativelysmall datasets of more or less homogeneous polymers and are generally not well val-idated. As such, they indicate that there may be useful chemometric methods here,but there is considerable scope for further studies on much larger and heterogeneoussample sets to demonstrate general applicability and usefulness.


3.2 Property Prediction

A more common use of informatics for data analysis is the development of(quantitative) structure-property relationships (QSPR) for the prediction of ma-terials properties and thus ultimately the design of polymers. Quantitativestructure-property relationships are multivariate statistical correlations betweenthe property of a polymer and a number of variables, which are either physicalproperties themselves or descriptors, which hold information about a polymer ina more abstract way. The simplest QSPR models are usually linear regression-type models but complex neural networks and numerous other machine-learningtechniques have also been used.

Two very simple types of QSPR have been developed early on in the evolutionof polymer property prediction, namely van Krevelen’s group contribution methods[122] and Bicerano’s system [123], which mainly relies on the use of topologicaldescriptors. Group contributions regard the overall properties of the polymer as thescalar sum of the properties of the chemical groups contained in the molecules mak-ing up the polymer.

While both the Bicerano and van Krevelen systems model a significant numberof polymer properties, most QSPR studies have focused on only a small numberof key properties (which is mainly correlated to the availability of data for modeldevelopment).

3.2.1 Glass Transition Temperature (Tg)

The glass transition temperature is a central property of polymers and of consid-erable importance in both fundamental polymer science as well as polymer engi-neering and processing. Below the glass transition temperature, macromolecules inbulk are fairly rigid, as they only have the freedom to vibrate and oscillate aroundfixed positions, creating a small amount of free volume. Significant translationalmovement, however, is usually not possible. The glass transition occurs at the tem-perature at which the free volume in the bulk material becomes large enough formacromolecular chains to move relative to each other. At this point the polymerbackbone relaxes and the material undergoes a transition from a solid to a quasiliq-uid state [124].

Van Krevelen’s group contributions are widely used for the prediction of Tg andperform reasonably well. When experimentally determined Tg values for 600 poly-mers are compared to predictions from group contributions, it could be shown thatapproximately 80% of the calculated Tg values were within±20K of the experimen-tal result [122]. A serious limitation of any group contribution method, however, isthat only polymers with structural groups for which contributions have been devel-oped can be predicted.

The group contribution methodology was extended by Hopfinger and Koehlerthrough combination with molecular modeling [125, 126]. In these papers themain determinants of the glass transition temperature are considered to be the

134 N. Adams

conformational entropy and mass moments of the polymer, which the authorsestimate from molecular mechanics and conformational energy calculations. Re-sults show, that Tg can be correlated with the intramolecular flexibility of thepolymer chain, which can be broken down into linear contributions of conforma-tional entropies of the repeat units and intermolecular interactions arising mainlyfrom electrostatics.

Katritzky and co-workers reported a descriptor-based attempt to develop QSPRmodels of Tg for low-molecular weight homopolymers [127]. The authors usedthe CODESSA suite [128] to calculate a set of 238 molecular descriptors for aset of 22 simple homo- and co-polymers with little structural diversity. To elimi-nate highly correlated descriptors, a pairwise comparison of descriptors was carriedout and only those with descriptors with pairwise correlation coefficients R2

ij < 0.1were used for the development of QSPR models. Linear regression techniques pro-duce a four parameter model with important contributions from descriptors such asDPSA (the difference between the positive and negative partial surface areas nor-malized by the number of electrons), the topological Randic index, the number ofOH groups present in the molecules, and the partial negative charge surface areaweighted by the total charge. The authors point out that molecules with large DSPAvalues have stronger intermolecular electrostatic interactions and thus higher glasstransition temperatures. The Randic index can be interpreted to be a measure ofbranching of the molecule. The regression model determined by the authors sug-gests, that the higher the degree of branching, the higher Tg, which is commensuratewith our current understanding of the glass transition temperature in polymers. Theappearance of the OH group count may be specific to the set of polymers chosen bythe researchers, but could also be interpreted as a measure of the hydrogen bond-ing in the system. The partial negative surface area, like the DSPA descriptor, maybe interpreted to be a measure of the electrostatic interactions between polymerchains. Unfortunately, the authors do not provide any type of validation of theirQSPR model, and therefore it is difficult to assess how general the model is.

A subsequent paper expands this work by considering a larger and more diverseset of polymers and by validating constructed models [129]. The work only con-siders linear polymers and uses trimers as input models. Descriptors were againcalculated using the CODESSA package on the middle fragment of the trimer. Lin-ear regression modeling gave rise to a five-parameter model. The most significantdescriptors were found to be the moment of inertia (measuring the mass distributionaround the principal axis of rotation), the Kier shape index (coding for the numberof skeletal atoms, molecular branching etc.), the most negative atomic charge in themolecule, the HSA/TSFA descriptor (measuring hydrogen bond forming ability),and the fractional positive partially charged surface area (describing electrostatic in-teractions between molecules), thus confirming the physical picture emerging fromthe first study. While the reported correlations are good (R2 = 0.946, R2

cv = 0.938for the best model) and additional cross-validation was also satisfactory, no valida-tion using a text/external data set (i.e., data not used for developing the regressionmodel) was reported, which again makes it somewhat difficult to assess the stabilityof the model. The standard error of prediction was 32.9 K.


Yet another expansion of this work was reported by Cao et al. who developed aset of descriptors based on the rotation of the polymer side chain, the bond countof the freely rotating part of the side chain, the polarizability effect index, and adescriptor containing information relating to hydrogen bonding [130]. The authorsapplied the descriptors to the polymer set reported by Katritzky and found goodcorrelations between predicted and observed values with a standard deviation of ap-proximately 21 K and an absolute average error of 15.30 K. While this is somewhatbetter than the results obtained by Katritzky, it is still a very significant error andunderlines the difficulty associated with the task of accurately predicting the glasstransition temperature for polymers.

An attempt to design polymers with specific properties, thus solving the “inverse”structure-property relationship problem, was presented by Reynolds in an elegantpiece of work [131]. In a first step, Reynolds constructed a library of 17 polymers,chosen from a larger collection of 112 diphenol/diacid condensation polymers,using a diversity search method in order to derive quantitative structure-propertyrelationships for Tg and the contact angle (CA). The QSPRs were then evaluatedagainst the remaining polymers in the large library and could be shown to be per-forming well (R2 = 0.89,RMS error = 7K (validation set)). In a subsequent step, thevalidated models were used to build focused libraries of new condensation polymerswith specific glass transition temperatures and contact angles.

Following on from Reynolds’ work, Brown and co-workers also developed a so-lution to the inverse QSPR problem using models built on the signature moleculardescriptor and targeting polymer properties such as Tg, heat capacity, density, mo-lar volume, and cohesive energies [132]. The “forwards” equations are comparableto other work discussed here in terms of predictive ability. The researchers subse-quently use their models to “design” polymers within a given property target rangeand to validate their approach by the “re-disovery” of Nylon-6,10 from an inverseQSPR model.

Apart from regression techniques, artificial neural networks (ANNs) have alsobeen investigated for the development of predictive systems for Tg. ANNs are math-ematical constructs, which are designed to mimic loosely information processing inthe brain and in particular the functions of neurons [133] and are commonly used instatistics to model complex and often non-linear relationships between data. Whilethere is a plethora of different neural network architectures, there are some commonfeatures. All neural networks have a number of interconnected processing nodes,which store knowledge by a dynamic response to external inputs and also makeinformation available for use [134]. Knowledge is contained in the weighted distri-bution between processing nodes and a learning algorithm is used to determine andchange the weights of the processing units during the learning process (Fig. 12).

In an early paper, Osguthorpe et al. investigated the use of ANNs of several dif-ferent architectures for the prediction of the glass transition temperature of linearhomopolymers, using descriptor values computed from the structure of the cor-responding monomers [135]. The problem with any prediction from a monomerstructure is, of course, that it ignores the history of a polymer, which can signifi-cantly influence its glass transition temperature as well as physical factors such as

136 N. Adams

Ij = ƒ[Σ w2jI2

w1jI1 wj1Ij

wj2Ij

wjNIjwNjIN

i=l

NWij Ii + qj

Fig. 12 The structure of a simple ANN together with an illustration of neuron microstructure(inlaid box)

Fig. 13 General chemicalstructure of theBrocchini-Kohn Library ofpolyarylates (R, Y = pointsof structural variability in thependent chain and thebackbone) [3]

HO CH2

C NHO

CHCH2RO

O

OH

HOC

YC

OH

O O+

O CH2

C NHO

CHCH2RO

O

O CY

O

CO

n

n=1,2

n=1,2

n

n

the molecular weight dependence of Tg and others. Consequently, the best result ob-tained by this approach gives RMS errors of 35 K in a validation set with maximumerrors as large as 130 K. Nevertheless, and probably as expected, the study showsthat the monomer does carry some information about the glass transition tempera-ture of the corresponding polymer.

In a subsequent study, Mattioni and Jurs compared prediction approaches, whichtake the structure of a monomer as an “input structure” vs those starting from the


polymer repeating unit structure [136]. In the first part of their study, numericaldescriptors for a set of 165 monomers were calculated and the set was subse-quently split into a training and test set. The best linear regression model was a10-component model with an RMS error of 25.87 K in the training set and 26.68 Kin the test set. The authors found that descriptors which encode molecular size andbranching were particularly relevant as was a descriptor counting the number of sin-gle bonds in the monomer molecule (which may hold information about the degreeof flexibility of the corresponding polymer backbone). The most relevant descriptorsfrom linear modeling were used as inputs into several NNs with different architec-tures. The best NN-generated models produced an RMS error of 15.67 K for thetraining and 21.76 K for the test set, which is an improvement when compared toOsguthorpe’s work. The use of genetic algorithms or simulated annealing proce-dures to select relevant input descriptors did not lead to a significant improvementin prediction accuracy of the NNs. In the second part of the study, descriptors de-rived from the repeating unit structure of a macromolecule were used. The bestlinear models constructed from this data had RMS errors of 40.06 K in the trainingand 43.16 K in the test set, which is significantly worse than the results achievedusing the monomer structures. The use of neural networks improves the situationsomewhat with RMS errors of 27.33 K and 32.96 K for training and test set. Formodels derived from repeating units, the use of descriptors selected using featureselection methods, in particular simulated annealing, as inputs led to significant im-provements in performance with the best NN models achieving RMS errors of 21.14and 21.94 for training and test sets respectively.

An unconventional approach to the prediction of Tg (other properties, such asdegradation temperature, tensile strength, Izod impact, Rockwell hardness, com-pressive strength, maximum elongation and refractive index were also considered)using neural networks was presented by Ulmer et al., who combined the use of“custom” descriptors for polymers with the notion of bootstrap cross-validation and“property experts” in an attempt to develop new polycarbonate polymers with de-fined physical properties [137]. The authors argue that traditional topological andtheoretical descriptors do not provide an optimum way to represent polymers, asthey tend to neglect aspects of polymer structure such as tacticity, entanglement,cross-linking, etc. They therefore introduce a new type of vector representation ofmolecular structure and atomic composition, which is referred to as “HamiltonianInteraction Modeling” (HIM). HIM is based on the assumption that the combinationof inter- and intramolecular interactions describes the behavior of both the atomisticand the condensed phase of a polymer. A “molecular Hamiltonian” for a repeatingunit is constructed by fusing descriptions of atomic, valence and mass connectivitywith an interaction site model based on cell models and the polymer reference in-teraction site model (PRISM) [138]. This yields several descriptors, which are usedas inputs into neural networks using “bootstrap resampling”. Furthermore, the au-thors compared the HIM descriptors to other descriptor systems, such as the Porter[139] and Bicerano [123] descriptors. Bootstrap resampling is used to subdivide thedataset into a training and test set. The algorithm works by creating a new datasetof N data points by randomly sampling an original dataset N times with the possi-

138 N. Adams

bility of sample repetition. Unique training sets are generated by comparing the setsproduced by bootstrap resampling to the original dataset and selecting only uniquedata points. Once a significant number of training and test sets have been createdin this way, neural networks are trained for each training set and evaluated againstan ensemble of test sets. Predicted values for the property under consideration arethen produced by averaging over the output of each of the individual networks.The trained networks were used to design novel bisphenol A polycarbonate (BPAC)polymers with increased impact resistance.

Afantitis et al. investigated the use of radial basis function (RBF) neural networksfor the prediction of Tg [140]. Radial basis functions are real-valued functions,whose value only depends on their distance from an origin. Using the dataset anddescriptors described in Cao’s work [130] (see above), RBF networks were trained.The best performing network models showed high correlations between predictedand experimental values. Unfortunately the authors do not formally report an RMSerror, but a cursory inspection of the reported data in the paper would suggest ap-proximate errors of around 10 K.

A slew of almost identical papers by a Chinese group reported the use of quantummechanical methods for the calculation of descriptors for several classes of poly-mers and their subsequent correlation to the glass transition temperature and otherpolymer properties via artificial neural networks [141–144]. The general conclu-sions, which can be drawn from these contributions are that NNs usually show betterperformance in predicting glass transition temperatures than regression models, andthat descriptors which codify for inter- and intramolecular interactions, conforma-tional freedom and the presence and size of a side chain, are the most suitable forpredicting Tg. This confirms results from prior studies using computationally lessexpensive methods.

Solaro et al. reported the use of a direct structure representation (chemical trees),rather than descriptors, as an input into a recursive neural network using a narrowdataset composed of methacrylate polymers containing alkyl, thioalkyl and fluo-roalkyl ester groups as well as polyacrylamides and α-substituted polyacrylics. Thisseems to lead to good prediction results in both training and test sets, though furthervalidation using larger and more diverse datasets will be required [145].

3.2.2 Refractive Index

The refractive index is another important polymer property and is defined as the ratioof the velocity of light traveling in a vacuum to that of light traveling in a material[123]. As polymers are increasingly used in optical applications, this quantity is ofconsiderable importance. The refractive index is also an important quantity in lightscattering experiments which, in turn, are used for the determination of molecularweights, molecular shapes, and molecular dimensions. As is the case for all majorpolymer properties, both the van Krevelen [122] and Bicerano [123] methodologiesallow the estimation of the quantity.


One of the first QSPR models was developed by Katritzky using a datasetconsisting of 95 mainly homochain polymers, though polyoxides, polyamides, andpolycarbonates were also represented [146]. The homochain polymers subdivideinto polyethylenes, polyacrylates, polymethacrylates, and polystyrenes. The modelwas developed from standard CODESSA descriptors using multilinear regression,which gave rise to a five-parameter equation (R2 = 0.940,s2 = 0.00013). The signif-icant parameters are the HOMO-LUMO energy gap, the AM1 heat of formation, themaximum nuclear repulsion for a C–H bond, the partial negative surface area, andthe relative number of fluorine atoms. The researchers interpret the presence of theAM1 heat of formation and the positive contribution this factor makes to the “loose-ness” of electrons in the molecule and therefore to a greater amount of freedomto interact with light. They note that less stable compounds have higher refractiveindices. The presence of the maximum nuclear repulsion energy for C–H bonds istaken to encode information containing the hybridization state of the carbon atomsin the molecule and thus presumably also holds information concerning how elec-tromagnetic radiation interacts with the electrons contained in the molecules. Thepartial negative surface area encodes information about the charge distribution andthus the size of the repeating unit. The presence of the fluorine atom count can be ra-tionalized by the very unusual electronic properties this atom imparts on molecules(fluorinated polymers have unusually low refractive indices).

Subsequent work was mainly reported by a cluster of Chinese workers whooutlined the development of a four-parameter model [147], the use of descriptorsderived from high-level density functional theory calculations [148], the use ofcyclic dimer structures rather than repeat unit structures for the purposes of de-scriptor calculation [149], and the development of a more specialized QSPR modelfor the refractive indices of conjugated polymers [150] and vinyl polymers [148].For the purposes of the development of their four-parameter model, Xu et al. usedmonomer structures rather than repeating unit structures. While they do not reportthe details of the software used to calculate their descriptors, four quantities seemto be of importance: the sum of the valence degrees, the degree of unsaturation, therelative number of halogen atoms, and the intermolecular electrostatic attraction orhydrogen bonding of the molecules. Conceptually, many of these descriptors en-code similar types of information as those used by Katritzky et al. and thus supportearlier findings. There is, however, only a marginal improvement of predictive accu-racy. The other papers in this series mainly confirm and repeat these results withoutadding significant new insights or improvements in predictive ability.

3.2.3 Lower Critical Solution Temperature (LCST)

The lower critical solution temperature is another crucial polymer property, which,together with the Upper Critical Solution Temperature (UCST), defines the twosolubility boundaries of polymers in solution. Typically, systems are completelymiscible below the LCST but only partially miscible above the LCST and com-pletely immiscible above the UCST.

140 N. Adams

Liu and Zhong introduced a number of QSPR models based on molecularconnectivity indices [151, 152]. In a first iteration, the researchers developedpolymer-dependent correlations: descriptors were calculated for a set of solventsand models were developed per polymer type [151]. Polymer classes underconsideration were polystyrene, polyethylene, poly-1-butene, poly-1-pentene,poly(4-methyl-1-pentene), polydimethylsiloxane, and polyisobutylene. As the au-thors fail to provide any validation for their models, it is difficult to asses theirpredictive power. In a subsequent iteration and general expansion of this study,mixed and therefore more general models based on the calculated connectivityindices of both solvent and polymers were developed. While it is unclear from thepaper which polymer representation was used for the calculation of the connectivityindices, the best regression model (eight parameter model) yields only acceptablepredictive power (R2 = 0.77, R2

cv = 0.77, s = 34.47 for the training set, R2 = 0.75for a test set) [152]. Using the same dataset, Afantitis and co-workers subsequentlyexpanded this work by expanding the descriptor space and providing more rig-orous validation [153]. The researchers produced a nine-parameter model, whichshows improvements over the equation put forward by Liu et al. (R2 = 0.8860,R2

cv = 0.8546 for the training set, R2 = 0.8738 for the test set). Xu et al. tested thedataset previously used by Liu and Afantitis using a set of 199 topological Dragondescriptors [154]. This leads to a linear 10-factor model which shows approxi-mately the same predictive power as that developed by Afantitis et al. (R2 = 0.8874,R2

cv = 0.8658, s = 24.57). In a further paper, Xu et al. investigated an even largerdescriptor set and the use of neural networks for the prediction of LCSTs [155]. Theresearchers showed that the development of an initial linear model on the basis ofthe enlarged descriptor space does not lead to significant improvements in predictiveability in comparison to earlier work, but that the use of neural networks can lead tofurther improvements in the predictive ability of a model (R2 = 0.9625, s = 13.43for the training and R2 = 0.9524 for the test set). The obvious drawbacks here arethe lack of interpretability and reproducibility of neural network models.

3.2.4 Intrinsic Viscosity

Like the lower critical solution temperature, the intrinsic viscosity of a polymer so-lution is dependent on both the nature of the polymer and that of the solvent. Theintrinsic viscosity contains information about the volume associated with a givenamount of polymer in dilute solution and thus encodes information about the con-formational properties of a polymer chain. The quantity is therefore of importancefor those engaged in both polymer synthesis and processing [123, 156]. Using thesame methodology the researchers exploited for the development of their LCSTmodel, Afantitis et al. report the construction of a linear eight-component model forthe intrinsic viscosity [157]. The correlation was developed using a dataset contain-ing 65 different polymer-solvent combinations and 10 different polymers. A totalof 30 physicochemical, topological, and structural descriptors were calculated andthe polymers were represented as their corresponding monomers for the purposes


of descriptor generation. As was the case with LCST, the prediction of the intrinsicviscosity is difficult: the best performing model showed R2 = 0.759, R2

cv = 0.601,and a root mean square (RMS) error of 34.67 for the training set (R2 = 0.751,RMS = 49.39, test set). Both the HOMO and LUMO values as well as the topo-logical index of the solvent and the principal moment of inertia along the x-axisof the solvent, together with the dipole length, Conolly molecular surface area, theLUMO energy, and the molecular weight of the polymer are significant factors, thusindicating that the intrinsic viscosity depends on the molecular weights of both thepolymer and that of the solvent, the polymer and solvent structure, the interactionsbetween polymer and solvent as well as the electronic properties of both polymerand solvent.

Gharagheizi subsequently expanded the work by Afantitis et al. by expandingthe descriptor space (calculating 1664 descriptors for both solvent and polymer)and investigating the use of radial basis function neural networks, but using the samedata set. Parameter selection through genetic-algorithm based multilinear regressionleads to a five-parameter model with an R2 = 0.8112 and R2

cv = 0.7714. The impor-tant descriptor types in this model are electrotopological state descriptors (polymer),information content descriptors (polymer), radial distribution function descriptors(polymer), as well as a weighted total accessibility index descriptor (solvent). Whenthese descriptors are used as inputs into a neural network, a further improvement inR2 can be observed.

3.2.5 Biomaterials

Polymer-based biomaterials are becoming increasingly important, whether they areused as medical supplies (pipes, catheters, bags), prostheses, or dental materials, orin a pharmaceutical context as drug conjugates [4, 7, 158–160], protein conjugates[6, 158, 159, 161], synthetic vectors [12, 14, 18, 162, 163], or as immuno-adjuvants[164, 165].

Early and prescient work to develop correlations between biological observablesand the physico-chemical properties of polymers were reported by Brocchini andKohn [3, 32, 166]. Prior to the development of the models, the authors had beenengaged in the combinatorial synthesis of a 112-member library of polyarylates,prepared through the condensation of diphenols with diacids (Fig. 12).

The library was evaluated with respect to a number of physicochemical propertiessuch as glass transition temperatures and contact angles, and a number of correla-tions were developed. Furthermore, the polymers in the library were screened withrespect to fibrinogen absorption and fibroblast proliferation on a thin polymer filmand good correlations between proliferation, contact angle, and the backbone archi-tecture of the polymer could be established: fibroblasts proliferate effectively whenseeded on polymers in which oxygen substitutions are present in the side chain andthe main chain and when the contact angle is large. However, proliferation decreasesupon increasing contact angle in the absence of oxygen substitution. More recently,Kohn et al. built on this work and reported the in silico design and preparation of

142 N. Adams

a polymethacrylate library. The library was screened against fibrinogen adsorptionand the attachment and growth of fibroblast-like NIH3T3 cells. Descriptors werecalculated on the polymer structures using the Molecular Operating Environment(MOE) program and models were developed using polynomial neural networks.While validation data are provided, the authors fail to communicate both the mod-els they developed and the precise nature of some of the descriptors involved inthe correlations (presumably for commercial reasons). This makes the comparisonand assessment of this work difficult though the authors assert that factors similarto those observed for the polyarylates, namely the hydrophilicity/hydrophobicity ofthe polymer and the presence of heteroatoms as well as measures of electrostaticbehavior, make a significant contribution to the models describing the biologicalresponses [167].

3.2.6 Other Polymer Properties

There have been isolated QSPR studies of a number of other polymer properties.These include the dielectric constant [144], the dielectric dissipation factor (tan δ)[168], the solubility parameter [169], the molar thermal decomposition function[170], the vitrification temperature of polyarylene oxides [171], and quantities relat-ing to molecularly imprinted polymers [172, 173]. The interested reader is referredto the literature for further information.

4 Summary and Conclusions

Informatics in the domain of polymer science is an exciting and multifaceted chal-lenge. Many problems remain unsolved: there are still significant issues in therepresentation of fuzzy materials such as polymers. The move away from theparadigm of the connection table is absolutely necessary, but also requires a rethinkof the traditional ways of information representation in chemistry. Some of the tech-nological solutions may be found in the realm of the semantic web, which is alreadyimpacting the way in which scholarly information and scientific data are communi-cated and transmitted. The main problem, however, is a cultural one: unlike in otherdisciplines, there is currently no culture of data sharing and data re-use in chemistryand much of the materials sciences. Furthermore, the current scholarly publicationprocess is profoundly dysfunctional – even scientists who wish to share data do notcurrently have the tools or the infrastructure to do so. All of this directly impedesthe development of polymer informatics. In spite of all of these obstacles, work totackle both the technological and cultural problems associated with polymer datais ongoing. Work is being carried out which attempts to relate polymer propertiesto their (chemical structure) representation. One should therefore remain optimistic,that progress towards the ultimate goal, namely the use of informatics to aid the un-derstanding of the physico-chemical behavior of polymers and their in silico designis inexorable.


References

1. Yoshida M, Langer R, Lendlein A et al. (2006) From advanced biomedical coatings to multi-functionalized biomaterials. Polym Rev 46:347–375

2. Dewez JL, Lhoest JB, Detrait E et al. (1998) Adhesion of mammalian cells to polymersurfaces: from physical chemistry of surfaces to selective adhesion on defined patterns. Bio-materials 19:1441–1445

3. Brocchini S, James K, Tangpasuthadol V et al. (1998) Structure-property coorrelations in acombinatorial library of degradable biomaterials. J Biomed Mat Res 42:66–75

4. Cuchelkar V, Kopecek J (2006) Polymer-drug conjugates. In: Uchegbu IF and SchaetzleinAG (ed) Polym Drug Deliv, CRC Press, Boca Raton

5. Torchilin VP (2006) Polymorphic micelles as pharmaceutical carriers. Polym Drug Deliv111–130

6. Haag R, Kratz F (2006) Polymer therapeutics: concepts and applications. Angew Chem IntEdn 45:1198–1215

7. Khandare J, Minko T (2006) Polymer-drug conjugates: progress in polymeric prodrugs. ProgPolym Sci 31:359–397

8. Way JL, Petrikovics I, Jiang J et al. (2001) Application of dendrimeric polymers as a drugcarrier in pharmacology. Abstracts of Papers, 221st ACS National Meeting, San Diego, CA,United States, April 1–5, 2001 IEC-316

9. Kataoka K, Kwon GS, Yokoyama M et al. (1993) Block copolymer micelles as vehicles fordrug delivery. J Contr Rel 24:119–132

10. Malmsten M (2006) Soft drug delivery systems. Soft Matter 2:760–76911. Qiu LY, Bae YH (2006) Polymer architecture and drug delivery. Pharm Res 23:1–3012. Kang HC, Lee M, Bae YH (2007) Polymeric gene delivery vectors. In: Peppas NA, Hilt JZ,

Thomas JB (ed) Nanotechnology in therapeutics Taylor and Francis, New York13. Alexis F, Zeng J, Wang S (2007) PEI nanoparticles for targeted gene delivery. Gene Transfer

473–47814. Leong KW (2006) Polymer design for nonviral gene delivery. BioMEMS Biomed Nanotech-

nol 1:239–26315. Mahato RI (2005) Water insoluble and soluble lipids for gene delivery. Adv Drug Deliv Rev

57:699–71216. Mahato RI, Kim SW (2005) Water soluble lipopolymers for gene delivery. In: Ammiji MM

(ed) Polym Gene Deliv, CRC Press, Boca Raton17. Adams ML, Lavasanifar A, Kwon GS (2003) Amphiphilic block copolymers for drug deliv-

ery. J Pharm Sci 92:1343–135518. Wagner E, Kloeckner J (2006) Gene delivery using polymer therapeutics. Adv Polym Sci

192:135–17319. Joester D, Losson M, Pugin R et al. (2003) Amphiphilic dendrimers: novel self-assembling

vectors for efficient gene delivery. Angew Chem Int Ed 42:1486–149020. Bjornerg HC, Derici L, Haggman BH et al. (2006) Hair care compositions comprising a

dendritic polymer. 2005-EP7017 200601806421. Derici L, Harcup JP, Khoshdel E (2006) Hair care composition comprising a dendritic macro-

molecule. 2005-EP7016 200601806322. Goosey M (2007) An overview of polymers as key enablers in electronics assembly-a

printed circuit board perspective. Polymers in Electronics 2007: Paper9/1-Paper9/5, Munich,Germany

23. Rost H (2007) Printed electronic circuits. Kunstst 97:97–10124. Xing R-b, Ding Y, Han Y-c (2007) Patterning of polymer by inkjet printing and its application

in the fabrication of organic electronic devices. Fenzi Kexue Xuebao 23:75–8125. Liang Z, Wang Q (2007) Patterning of conjugated polymers for organic electronics and opto-

electronics. In: Naiwa HS (ed) Polym Nanostruct Their Appl, American Scientific Publishers,Stevenson Ranch, California

144 N. Adams

26. Bock K (2005) Polytronics – electronics and systems on flexible substrates. IEEE VLSI-TSAInternational Symposium on VLSI Technology, Hsinchu, Taiwan, pp 53–56

27. Stafford N (2007) Large-scale biopolymer production. http://www.rsc.org/chemistryworld/News/2007/May/14050701.asp, Accessed Dec 12 2008

28. Zhang H, Hoogenboom R, Meier MAR et al. (2004) High-throughput experimentation inpolymer chemistry. Trans Mater Res Soc Jpn 29:319–324

29. Zhang H, Hoogenboom R, Meier MAR et al. (2005) Combinatorial and high-throughput ap-proaches in polymer science. Meas Sci Technol 16:203–211

30. Hoogenboom R, Fijten MWM, Wijnans S et al. (2006) High-throughput synthesis and screen-ing of a library of random and gradient copoly(2-oxazoline)s. J Comb Chem 8:145–148

31. Hoogenboom R, Schubert US (2005) High-throughput synthesis equipment applied to poly-mer research. Review of Scientific Instruments 76:062202/062201–062202/062207

32. Brocchini S, James K, Tangpasuthadol V et al. (1997) A combinatorial approach for polymerdesign. J Am Chem Soc 119:4553

33. Wiesbrock F, Hoogenboom R, Leenen MAM et al. (2005) Investigation of the living cationicring-opening polymerization of 2-methyl-, 2-ethyl-, 2-nonyl-, and 2-phenyl-2-oxazoline in asingle-mode microwave reactor. Macromolecules 38:5025–5034

34. Wiesbrock F, Hoogenboom R, Abeln CH et al. (2004) Single-mode microwave ovens as newreaction devices: accelerating the living polymerization of 2-ethyl-2-Oxazoline. MacromolRapid Commun 25:1895–1899

35. Gilman JW, Bourbigot S, Shields JR et al. (2003) High throughput methods for polymernanocomposites research: extrusion, NMR characterization and flammability property screen-ing. J Mat Sci 38:4451

36. Davis RD, Bur AJ, McBearty M et al. (2004) Dielectric spectroscopy during extrusion pro-cessing of polymer nanocomposites: a high-throughput processing/characterization methodto measure layered silicate content and exfoliation. Polymer 45:6487–6493

37. Gilman JW, Davis RD, Bellayer S et al. (2005) Use of optical probes and laser scanning con-focal fluorescence microscopy for high-throughput characterization of dispersion in polymerlayered silicate nanocomposites. PMSE Prepr 92:168–169

38. Gilman JW, Davis RD, Shields JR et al. (2004) Development of high-throughput methodsfor polymer flammability property characterization. International SAMPE Symposium andExhibition:460–469

39. Gilman JW, Maupin PH, Harris RH et al. (2004) High throughput methods for nanocompositematerials research. Extrusion and visible optical probes. PMSE Prepr. 90:717–718

40. Adams N, Moneke M, Gulmus SA et al. (2006) Combinatorial compounding. Mater Res SocSymp Proc 894:171–179

41. Kranenburg JM, Tweedie CA, Hoogenboom R et al. (2007) Elastic moduli for a di-block copoly(2-oxazoline) library obtained by high-throughput screening. J Mater Chem17:2713–2721

42. Kranenburg JM, van Duin M, Schubert US (2007) Screening of EPDM cure states usingdepth-sensing indentation. Macromol Chem Phys 208:915–923

43. Cheung K-H, Yip KY, Townsend JP et al. (2008) HCLS 2.0/3.0: health care and life sciencesdata mashup using Web 2.0/3.0. J Biomed Inform 41:694–705

44. Walkingshaw AD, White TOH, Day NE et al. (2008) Representing, indexing and miningscientific data with XML and RDF: Golem and CrystalEye. XTech 2008: Dublin, Ireland

45. Ma H, Melillo G, Oliva L et al. (2005) Aluminum alkyl complexes supported by [OSSO] typebisphenolato ligands: synthesis, characterization and living polymerization of rac-lactide.Dalton Trans 721–727

46. Huggins ML (1969) Macromolecular nomenclature: general background and perspective.J Chem Doc 9:230–231

47. Livingston HK, Fox RB (1969) Nomenclature of organic polymers. J Chem Doc 9:232–23448. Cohn WE (1969) Representation of macromolecules and polymers of biological importance.

J Chem Doc 9:235–24149. Block BP, Thomas PM, Donovan KM (1969) Problems in the nomenclature of inorganic

polymers. J Chem Doc 9:242–244


50. Bikales NM (1969) Polymer nomenclature in industry. J Chem Doc 9:245–24751. Loening KL, Metanomski WV, Powell WH (1969) Indexing of polymers in Chemical

Abstracts. J Chem Doc 9:248–25152. Metanomski WV (1979) Symposium on retrieval of polymer information: introductory

remarks. J Chem Inf Comput Sci 19:5953. Langstaff EM, Ostrum K (1979) Access to polymer information in chemical abstracts. J Chem

Inf Comput Sci 19:60–6454. Fugmann R (1979) POLIDCASYR: the polymer documentation system of IDC. J Chem Inf

Comp Sci 19:64–6855. Donaruma LG (1979) Some problems encountered in interdisciplinary searches of the poly-

mer literature. J Chem Inf Comp Sci 19:68–7056. Nardone J (1979) Computerized numeric data for polymers. J Chem Inf Comp Sci 19:71–7357. Roush PF, Seitz JT, Young LF (1979) An on-line system for storage and retrieval of polymer

data. J Chem Inf Comp Sci 19:73–7658. Skolnik H (1979) A classification system for polymer literature in an industrial environment.

J Chem Inf Comp Sci 19:76–7959. Zurbach Balent M, Lotz JW (1979) Polymers and patents don’t mix-easily. J Chem Inf Comp

Sci 19:80–8360. Fugmann R (1974) Representation of concept relations using the TOSAR system of the IDC.

J Am Soc Inf Sci 25:287–30761. Manola F, Miller E (2004) RDF Primer. http://www.w3.org/TR/rdf-primer/. Accessed Jul 10

200762. Brickley D, Guha RV (2004) RDF vocabulary description language 1.0: RDF schema.

http://www.w3.org/TR/rdf-schema/. Accessed Dec 30 200863. McGuiness D, van Harmelen F (2004) OWL web ontology language overview. http:-

//www.w3.org/TR/owl-features/.Accessed Dec 30 200864. Ranganathan SR (1963) Colon classification. Asia Publishing House, Bombay, India65. Metanomski WV (1991) Compendium of macromolecular nomenclature (the purple book).

Blackwell Scientific Publications, Oxford66. Kaback SM (1991) Polymer information: storage for retrieval, or hide and seek? Introduction.

J Chem Inf Comput Sci 31:439–44367. Gushurst AJ, Nourse JG, Hounshell WD et al. (1991) The substance module: the representa-

tion, storage and searching of complex structures. J Chem Inf Comp Sci 31:447–45468. Kaback SM (1991) There’s more to a polymer than just its build. J Chem Inf Comput Sci

31:439–44369. Briggs JA, Ferns EA, Shenton KE (1991) Improvements in Derwent Plasdoc system. J Chem

Inf Comput Sci 31:454–45870. Rieder MD (1991) The IFI polymer indexing system: its past, present and future. J Chem Inf

Comput Sci 31:458–46271. Green C (1991) The Rapra abstracts rubber and plastics database. J Chem Inf Comput Sci

31:476–48172. Herz M (1991) Polymer searching in different databases. J Chem Inf Comput Sci 31:469–47573. Lambert N (1991) Online searching of polymer patents: precision and recall. J Chem Inf

Comput Sci 31:443–44674. Wilke RN, Buntrock RE (1991) Condensation polymer information: problems and opportu-

nities. J Chem Inf Comput Sci 31:463–46875. Berners-Lee T, Hendler J, Lassila O (2001) The semantic web. Sci Am 284:34–4476. Bray T, Paoli J, Sperberg-McQueen CM et al. (2006) Extensible markup language (XML) 1.1

(Second Edition). http://www.w3.org/TR/REC-xml/.Accessed Jul 10 200777. W3C (2004) XML schema part 0: primer. Second edition http://www.w3.org/TR/xmlschema-

0/. Accessed Dec 12 200878. Holliday GL, Murray-Rust P, Rzepa HS (2006) Chemical markup, XML, and the world

wide web. 6. CMLReact, an XML vocabulary for chemical reactions. J Chem Inf Model46:145–157

146 N. Adams

79. Murray-Rust P, Rzepa HS, Williamson MJ et al. (2004) Chemical markup, XML, and theworld wide web. 5. Applications of chemical metadata in RSS aggregators. J Chem Inf Com-put Sci 44:462–469

80. Murray-Rust P, Rzepa HS (2003) Chemical markup, XML, and the world wide web. 4. CMLschema. J Chem Inf Comput Sci 43:757–772

81. Gkoutos GV, Murray-Rust P, Rzepa HS et al. (2001) Chemical markup, XML and theworld-wide web. 3. Toward a signed semantic chemical web of trust. J Chem Inf ComputSci 41:1124–1130

82. Murray-Rust P, Rzepa HS (2001) Chemical markup, XML and the world-wide web. 2. Infor-mation objects and the CMLDOM. J Chem Inf Comput Sci 41:1113–1123

83. Murray-Rust P, Rzepa H (1999) Chemical markup, XML, and the world-wide web. 1. Basicprinciples. J Chem Inf Comput Sci 39:928–942

84. Frenkel M, Chiroco RD, Diky V et al. (2006) XML-based IUPAC standard for experimental,predicted, and critically evaluated thermodynamic property data storage and capture (Ther-moML) (IUPAC Recommendations 2006). Pure Appl Chem 78:541–612

85. Sankar P, Aghila G (2006) Design and development of chemical ontologies for reaction rep-resentation. J Chem Inf Model 46:2355–2368

86. Sankar P, Aghila G (2007) Ontology aided modeling of organic reaction mechanisms withflexible and fragment based XML markup procedures. J Chem Inf Model 47:1747–1762

87. Microsoft (2008) Chem4Word project. http://research.microsoft.com/projects/chem4word/.Accessed Dec 30 2008

88. Adams N, Murray-Rust P (2008) Engineering polymer informatics: towards the computer-aided design of polymers. Macromol Rapid Commun 29:615–632

89. Adams N, Murray-Rust P, Winter J et al. (2008) Chemical markup, XML and the world wideweb. 8. Polymer Markup Language. J Chem Inf Model 48:2118–2128

90. Clark J (1999) XSL Transformations (XSLT). http://www.w3.org/TR/xslt. Accessed Aug 042008

91. de Matos P, Ennis M, Zbinden M et al. (2006) ChEBI – Chemical entities of biological inter-est. http://www3.oup.co.uk/nar/database/summary/646, Accessed Dec 12 2008

92. Kanehisa M, Goto S, Kawashima S et al. (2004) The KEGG resource for decipering thegenome. Nucleic Acids Res 32:D277–D280

93. Fleischmann A, Darsow M, Degtyarenko K et al. (2004) IntEnz, the integrated relationalenzyme database. Nucleic Acids Res 32:D434–D437

94. Degtyarenko K (2007) The Rex ontology. http://obofoundry.org/cgi-bin/detail.cgi?id=rex,Accessed Dec 30 2008

95. Degtyarenko K (2007) The FIX ontology. http://obofoundry.org/cgi-bin/detail.cgi?id=fix,Accessed Dec 30 2008

96. Feldman HJ, Dumontier M, Lng S et al. (2005) CO: a chemical ontology for identification offunctional groups and semantic comparison of small molecules. FEBS Lett 579:4685–4691

97. Frey JG, Hughes GV, Mills HR et al. (2003) Less is more: lightweight ontologies and userinterfaces for smart labs. UK e-Science All Hands Meeting:500–507, Nottingham, UK

98. Frey JG, de Roure D, Schraefel MC et al. (2003) Context slicing the chemical aether. FirstInternational Workshop on Hypermedia and the Semantic Web:9, Nottingham, UK

99. Taylor KR, Gledhill RJ, Essex JW et al. (2006) Bringing chemical data onto the semanticweb. J Chem Inf Model 46:939–952

100. Soldatova LN, Clare A, Sparkes A et al. (2006) An ontology for a robot scientist. Bioinfor-matics 22:e464–e471

101. Niles I, Pease A (2001) Towards a standard upper ontology. Proceedings of the 2nd Inter-national Conference on Formal Ontology in Information Systems (FOIS-2001): Ogunquit,Maine, United States

102. Heller B, Herre H (2004) Ontological categories in GOL. Axiomathes 14:57–76103. Brandrup J, Immergut EH (1989) Polymer handbook. Wiley, New York104. King J (2008) Text content in pdf files. http://blogs.adobe.com/insidepdf/2008/07/text_content

_in_pdf_files.html. Accessed Dec 28 2008


105. Fenniri H, Chun S, Terreau O et al. (2007) Preparation and infrared/Raman classification of630 spectroscopically encoded styrene copolymers. J Comb Chem 10:31–36

106. Hall SR, Allen FH, Brown ID (1991) The Crystallographic Information File (CIF): a newstandard archive file for crystallography. Acta Cryst A 47:655–685

107. CCDC (2008) The Cambridge Crystallographic Data Centre. http://www.ccdc.cam.ac.uk/.Accessed Dec 12 2008

108. Day NE (2008) CrystalEye. http://wwmm.ch.cam.ac.uk/crystaleye/index.html. Accessed Dec12 2008

109. Corbett P, Murray-Rust P (2006) High-throughput identification of chemistry in life sciencetexts. Computational Life Sciences II. Lecture Notes in Computer Science, vol 4216, pp107–118

110. Atkinson D (1992) The evolution of medical research writing from 1735 to 1985: the case ofthe Edinburgh Medical Journal. Appl Linguist 13:337–374

111. Zaye DF, Metanomski WV (1986) Scientific communication pathways: an overview and in-troduction to a symposium. J Chem Inf Comput Sci 26:43–44

112. Suh C, Rajagopalan A, Li X et al. (2002) The application of principal component analysis tomaterials science data. Data Sci J 1:19

113. Bajorath J (2001) Selected concepts and investigations in compound classification, moleculardescriptor analysis and virtual screening. J Chem Inf Comput Sci 41:233

114. Wold S, Esbensen K, Geladi P (1987) Principal component analysis. Chemom Intell Lab Syst2:37

115. Vanden Eynde X, Bertrand P (1997) ToF-SIMS quantification of polystyrene spectra basedon principal component analysis (PCA). Surf Interface Anal 25:878

116. Coullerez G, Lundmark S, Malmstroem E et al. (2003) ToF-SIMS for the characterizationof hyperbranched aliphatic polyesters: probing their molecular weight on surfaces based onprincipal component analysis (PCA). Surf Interface Anal 35:693–708

117. Batur C, Vhora MH, Cakmak M et al. (1999) On-line crystallinity measurement using laserRaman spectrometer and neural network. ISA Trans 38:139–148

118. Miranda TMR, Goncalves AR, Amorim MTP (2001) Ultraviolet-induced crosslinking ofpoly(vinyl alcohol) evaluated by principal component analysis of FTIR spectra. Polym Int50:1068–1072

119. Vazquez C, Boeykens S, Bonadeo H (2002) Total reflection X-ray fluorescence polymer spec-tra: classification by taxonomy statistic tools. Talanta 57:1113–1117

120. Tuchbreiter A, Marquardt J, Zimmermann J et al. (2001) High-throughput evaluation of olefincopolymer composition by means of attenuated total reflection fourier transform infraredspectroscopy. J Comb Chem 3:598–603

121. Lukasiak BM, Faria R, Zomer S et al. (2006) Pattern recognition for the analysis of polymericmaterials. Analyst 131:73–80

122. van Krevelen DW (1990) Properties of polymers: their correlation with chemical struc-ture, their numerical estimation and prediction from additive group contributions. Elsevier,Amsterdam

123. Bicerano J (2002) Prediction of polymer properties. Marcel Dekker Ltd, New York124. Stevens MP (1990) Polymer chemistry. An introduction. Oxford University Press, Oxford125. Koehler MG, Hopfinger AJ (1989) Molecular modelling of polymers: 5. Inclusion of inter-

molecular energetics in estimating glass and crystal-melt transition temperatures. Polymer30:116–126

126. Hopfinger AJ, Koehler MG, Pearlstein RA (1988) Molecular modling of polymers. IV. Esti-mation of glass transition temperatures. J Polym Sci Part B 26:2007–2028

127. Katritzky AR, Rachwal P, Law KW et al. (1996) Prediction of polymer glass transition tem-peratures using a general quantitative structure-property relationship treatment. J Chem InfComput Sci 36:879–884

128. Ivanciuc O (1997) CODESSA version 2.13 for Windows. J Chem Inf Comput Sci 37:405–406129. Katritzky AR, Sild S, Lobanov V et al. (1998) Quantitative structure-property relationship

(QSPR) correlation of glass transition temperatures of high molecular weight polymers.J Chem Inf Comput Sci 38:300–304

148 N. Adams

130. Cao C, Lin Y (2003) Correlation between the glass transition temperatures and repeating unitstructure for high molecular weight polymers. J Chem Inf Comput Sci 43:643–650


132. Brown WM, Martin S, Rintoul MD et al. (2006) Designing novel polymers with targetedproperties using the signature molecular descriptor. J Chem Inf Model 46:826–835

133. Gurney K (1997) An introduction to neural networks. Routledge, London134. Sumpter BG, Getino C, Noid DI (1994) Theory and applications of neural computing in

chemical science. Annu Rev Phys Chem 45:439–481135. Joyce SJ, Osguthorpe DJ, Padgett JA et al. (1995) Neural network prediction of glass-

transition temperatures from monomer structure. J Chem Soc Faraday Trans 91:2491–2496136. Mattioni BE, Jurs PC (2002) Prediction of glass transition temperatures from monomer and

repeat unit structure using computational neural networks. J Chem Inf Comput Sci 42:232–240

137. Ulmer II CW, Smith DA, Sumpter BG et al. (1998) Computational neural networks and therational design of polymeric materials: the next generation polycarbonates. Comput TheorPolym Sci 8:311–321

138. Schweizer KS, Curro JG (1994) PRISM theory of the structure, thermodynamics, and phasetransitions of polymer liquids and alloys. Adv Polym Sci 116:319–377

139. Porter D (1995) Group interaction modeling of polymer properties. Marcel Dekker, New York140. Afantitis A, Melagraki G, Makridima K et al. (2005) Prediction of high weight polymers glass

transition temperature using RBF neural networks. J Mol Struct: THEOCHEM 716:192–198141. Yu X, Yi B, Wang X et al. (2007) Correlation between the glass transition temperatures and

multipole moments for polymers. Chem Phys 332:115–118142. Gao J, Wang X, Li X et al. (2006) Prediction of polyamide properties using quantum-chemical

methods and BP artificial neural networks. J Mol Model 12:513–520143. Liu W, Yi P, Tang Z (2006) QSPR Models for various proeprties of polymethacrylates based

on quantum chemical descriptors. QSAR Comb Sci 25:936–943144. Liu A, Wang X, Wang L et al. (2007) Prediction of dielectric constants and glass transi-

tion temperatures of polymers by quantitative structure-property relationships. Eur Polym J43:989–995

145. Duce C, Michell A, Starita A et al. (2006) Prediction of polymer properties from their struc-ture by recursive neural networks. Macromol Rapid Commun 27:711–715

146. Katritzky AR, Sild S, Karelson M (1998) Correlation and prediction of the refractive indicesof polymers by QSPR. J Chem Inf Comput Sci 38:1171–1176

147. Xu J, Chen B, Zhang Q et al. (2004) Prediction of refractive indices of linear polymers by afour descriptor QSPR model. Polymer 45:8651–8659

148. Yu X, Yi B, Wang X (2007) Prediction of the refractive index of vinyl polymers by usingdensity functional theory. J Comp Chem 28:2336–2341

149. Xu J, Liang H, Chen B et al. (2008) Linear and nonlinear QSPR models to predict refractiveindices of polymers from cyclic dimer structures. Chemom Intell Lab Syst 92:152–156

150. Gao J, Xu J, Chen B et al. (2007) A quantitative structure-property relationship study forrefractive indices of conjugated polymers. J Mol Model 13:573–578

151. Liu H, Zhong C (2005) Modeling of the theta (lower critical solution temperature) in polymersolutions using molecular connectivity indices. Eur Polym J 41:139–147

152. Liu H, Zhong C (2005) General correlation for the prediction of theta (lower critical solutiontemperature) in polymer solutions. Ind Eng Chem Res 44:634–638

153. Melagraki G, Afantitis A, Sarimveis H et al. (2007) A novel QSPR model for predicting theta(lower critical solution temperature) in polymer solutions using molecular descriptors. J MolModel 15:55–64

154. Xu J, Liu L, Xu W et al. (2007) A general QSPR model for the prediction of theta (lowercritical solution temperature) in polymer solutions with topological indices. J Mol GraphModel 26:352–359


155. Xu J, Chen B, Liang H (2008) Accurate prediction of theta (lower critical solutiontemperature) in polymer solutions based in 3D descriptors and artificial neural networks.Macromol Theory Simul 17:109–120

156. Rushing TS, Hester RD (2004) Semi-empirical model for polyelectrolyte intrinsic viscosityas a function of ionic strength and polymer molecular weight. Polymer 45:6587–6594

157. Afantitis A, Melagraki G, Sarimveis H et al. (2006) Prediction of intrinsic viscosity inpolymer-solvent combinations using a QSPR model. Polymer 47:3240–3248

158. Duncan R (2006) Polymer conjugates as anticancer nanomedicines. Nat Rev Cancer6:688–701

159. Duncan R, Ringsdorf H, Satchi-Fainaro R (2006) Polymer therapeutics: polymers as drugs,drug and protein conjugates and gene delivery systems: past, present and future opportunities.Adv Polym Sci 192:1–8

160. G.S. Kwon, K. Kataoka (1995) Block copolymer micelles as long-circulating drug vehicles.Adv Drug Delivery Rev 16:295

161. Hoffman AS, Stayton PS (2004) Bioconjugates of smart polymers and proteins: synthesis andapplications. Macromol Symp 207:139–151

162. Putnam D (2006) Polymers for gene delivery across length scales. Nat Mater 5:439–451163. Godbey WT, Wu KK, Mikos AG (1999) Poly(ethylenimine) and its role in gene delivery.

J Controlled Release 60:149–160164. Hunter R, Strickland F, Kezdy F (1981) The adjuvant activity of nonionic block polymer

surfactants. J Immunol 127:1244–1250165. Hunter RL, Bennett B (1984) The adjuvant activity of nonionic block polymer surfactants. II.

Antibody formation and inflammation related to the structure of the triblock and octablockcopolymer. J Immunol 133:3167–3175

166. Brocchini S (2001) Combinatorial chemistry and biomedical polymer development. AdvDrug Delivery Rev 53:123–130

167. Kholodovych V, Gubskaya A, Bohrer M et al. (2008) Prediction of biological response forlarge combinatorial libraris of biodegradable polymers: polymethacrylates as a test case.Polymer 49:2435–2439

168. Yu X, Yi B, Liu F et al. (2008) Prediction of the dielectric dissipation factor tan delta ofpolymers with an ANN model based on DFT calculation. React Funct Polym 68:1557–1562

169. Yu X, Wang X, Wang H et al. (2006) Prediction of solubility parameters for polymers by aQSPR model. QSAR Comb Sci 25:156–161

170. Yu X, Xie Z, Yi B et al. (2007) Prediction of the thermal decomposition property of polymersusing quantum chemical descriptors. Eur Polym J 818–823

171. Toropov AA, Nurgaliev IN, Balakhonenko OI et al. (2004) QSPR modeling of vitrificationtemperatures for polyarylene oxides. J Struct Chem 45:706–712

172. Nantasenamat C, Isarankura-Na-Ayudhya I, Naenna T et al. (2007) Quantitative structure-imprinting factor relationship of molecularly imprinted polymers. Biosens Bioelectron2007:3309–3317

173. Si HZ, Zhang KJ, Hu ZD et al. (2007) QSAR model for prediction capacity factor of molec-ular imprinting polymer based on gene expression programming. QSAR Comb Sci 26:41–50

174. Hamoudeh M, Faraj AA, Canet-Soulas E et al. (2007) Elaboration of PLLA-based superpara-magnetic nanoparticles: characterization, magnetic behaviour study and in vitro relaxivityevaluation. Int J Pharm 338:248–257

175. Service CA (1997) Chemical Abstracts Index Guide 1997. Columbus

Index

AAccess-barriers, 125Acrylate-based coatings, 10Acrylates, photopolymerization, 4N-Acryloyl morpholine (Amor), 27, 41Adhesion testing, 88Alkoxyamine, 262-Alkyl-2-oxazoline, 33Analytical Markup Language (AnIML), 122Artificial neural networks (ANNs), 135Atom transfer radical polymerization (ATRP),

22Automated parallel synthesis, automated, 17

BBakelite, 113Biomaterials, 141Bis(terpyridine) ruthenium, 53Blend gradient film, 4Block copolymers, 17, 44Bootstrap resampling, 137(1-Bromo ethyl) benzene (BEB), 22Brushes, 76Butyl acrylate, 10, 26Butyl methacrylate, 10, 30sec-Butyllithium (s-BuLi), 32

CCandida antarctica lipase, 8Cationic ring opening polymerization (CROP),

33Chemical cleaning, 32Chemical Markup Language (CML), 122Chemometrics, 130Chemspeed automated synthesizer, 6Coating libraries, 10Combinatorial materials research, 1Continuous gradient library techniques, 65

Controlled radical polymerization (CRP), 21Controlled/“living” polymerization (CLP), 20Copolymer libraries, 35Copoly(2-oxazoline)s, 502-Cyano-2-butyl dithio benzoate (CBDB), 30

DDegradable polymers, 9Dendrimers, 10N,N-Dimethyl acrylamide (DMA), 27N,N-Dimethyl aminoethyl acrylamide

(DMAEMA), 37Diphenols, tyrosine-derived, 7DNA-complexing materials, 8DT-A, 9DynamicMechanical Analysis (DMA), 132

EEdge-delamination tests, 91Elastic materials, 921-Ethoxy ethyl acrylate (EEA), 44Ethyl-2-bromo-iso-butyrate (EBIB), 222-Ethyl-2-oxazoline (EtOx), CROP, 34

FFilm thickness, 86Flow coating, 66Fouling-release potential, 10FT-IR, 131

GGene delivery vectors, 8Glass transition temperature, 7, 40, 51, 110,

133Glassy materials, 91

151

152 Index

Gradient hot stage, 68Gradient library, 66Gradient polymer brush libraries, 76

HHamiltonian Interaction Modeling (HIM), 137High-throughput, 1, 17High-throughput materials, synthesis, 94

– testing, 842-Hydroxyethyl acrylate, 102-Hydroxypropyl acrylate (HPA), 27, 412-Hydroxypropyl methacrylate (HPMA), 95

IInformation systems, 107Iniferter, 21Interfaces, 63, 84Interfacially-active polymers, microchannels,

96Intrinsic viscosity, 140Ionic polymerization, 32, 46N-Isopropyl acrylamide (NIPAM), 37

JJKR adhesion tests, 92

KKelen–Tüdös (KT) method, 48

LLDPE, 131Library preparation, 1Lipase, 8LLDPE, 109Lower critical solution temperature (LCST),

139

MMachine learning, 107Mayo-Lewis terminal model (MLTD), 48Methacrylic acid (MAA), 372-Methoxyethyl-2-methylacrylate (MeOMA),

37Methyl bromo propionate (MBP), 22Methyl methacrylate (MMA), ATRP, 22Microchannels, controlled polymer synthesis,

95Microfluidic devices, 94Microscale approaches, 94

Molecular Operating Environment (MOE),142

Monomer conversion, 24Multiarm star polymers, 10

NName-based representations, 117Near edge X-ray absorption fine structure

(NEXAFS), 78Nitroxide mediated polymerization (NMP), 22,

262-Nonyl-2-oxazoline (NonOx), 34Nylon-6,10, 135

On-Octyldimethylchlorosilane (ODS), 70Oligo(ethylene glycol) methacrylates, 37Oligo(ethylene glycol) methyl ether

methacrylate (OEGMA), 37Ontology, 1072-Oxazolines, 33

PP(Amor)-stat-(HPA), 40P(DMA-stat-HPA), 43P(EtOx)-stat-(SoyOx), 49P(St)-b-(t-BA), 46Peel test, 89Phase behavior, 42-Phenyl-2-oxazolines, 34PMMA, 42POLIDCASYR, 114Poly(acrylic acid) (PAA), 44Poly(β-amino esters), 5, 10Poly(arylene ethynylene)s, 9Poly-1,3-butadiene, 118Poly-1-butene, 140Poly(n-butyl acrylate) (PnBA), 44Poly(tert-butyl acrylate) (PtBA), 45Poly(n-butyl methacrylate) (PnBMA), 78Poly(N,N-dimethyl aminoethyl methacrylate)

(PDMAEMA), 44, 78Poly(1-ethoxyethyl acrylate) (PEEA), 44Poly(ethylene glycol) (PEG), 37Poly(ethylene oxide) (PEO), 57Poly(ethylene terephthalate) (PET), 117Poly(ethyleneimine), 9Poly(HEMA), 88Poly(methyl acrylate) (PMA), 44Poly(4-methyl-1-pentene), 140Poly-1-pentene, 140Poly(vinyl alcohol) (PVA), 131

Index 153

Polyanhydride random copolymers, 8Polyarylates, Brocchini-Kohn Library, 136Polybutadiene, 117Polydimethylsiloxane, 140Polydispersity index (PDI), 24, 112Polyethylene, 140Polyisobutylene, 140Polyisoprene-b-PS-b-poly(ethylene oxide), 74Polymer blend composition gradients, 82Polymer brush composition gradients,

surface-grafted, 5Polymer film thickness gradients, 66Polymer informatics, 107Polymer libraries, 1, 127

– fabrication, fluids, 94Polymer markup language, 107Polymer reference interaction site model

(PRISM), 137Polyols, 10Polystyrenes, 26, 140

– library, 127Polystyrene/polyethylene oxide

block-copolymer, 118Pressure sensitive adhesives (PSAs), 88Principal component analysis (PCA), 130Principal component regression (PCR), 132Probe tack tests, 90Property screening, 12-iso-Propyl-2-oxazoline (iPrOx), 37PS-b-PMMA, 73Pseudo-barnacle adhesion, 10

QQuantitative structure-property relationships

(QSPR), 107, 133

RRadial basis function (RBF) neural networks,

138Radical polymerization, 21, 35Random copolymers, 17, 47RDF, 107Refractive index, 138Resource Description Framework (RDF), 121Reversible addition fragmentation chain

transfer (RAFT), 22, 28

SSelf-assembled monolayer (SAM), 70Semantic web, 107, 120

Sgroups, 118Siloxane-polycaprolactone block copolymers,

6Siloxane-polyurethane coatings, 10Small angle light scattering (SALS), 131Source-based representations, 1172-“Soyalkyl”-2-oxazoline (SoyOx), 49Structure-based representations, 117, 118Structure–property relationships, 1Styrene, 21, 33Substrate-block interactions, 72Supramolecular synthesis, 53Surface chemistry libraries, 70Surface energy libraries, 70Surface-grafted copolymers, 4Surface-initiated polymerization (SIP), 76Symyx batch polymerization system, 6

TTemperature processing libraries, 68TEMPO, 26ThermoML (markup language for

thermochemical/-physicaldata), 122

Thermoset polymers, 7Thin films, crazing, 85

– mechanical properties, 84– coatings, modulus, 86

ToF-SIMS, 131p-Toluene sulfonyl chloride (TsCl), 22TOSAR, 115Total X-Ray Fluorescence (TXRF), 132

UUpper critical solution temperature (UCST),

139UV–ozone, 70

VViscoelastic materials, 89, 90

WWeb ontology language (OWL), 122

XXanthate RAFT agent, 28

Date post:	31-May-2020
Category:	Documents
Upload:	others
View:	2 times
Download:	0 times

mkimia.fst.unair.ac.idmkimia.fst.unair.ac.id/wp-content/uploads/2018/04/polymer-libraries.pdfAdvances...

Documents